Legislatures often instruct judges to impose harsher punishments on people who have prior criminal convictions—for example, a conviction for a “crime of violence” or for a “crime involving moral turpitude.” But how are judges to determine whether a person has such a conviction? In Mathis v. United States, the Supreme Court clarified that judges can rely on only the legal “elements” of prior convictions, not the factual “means” the person employed in committing the prior offense. The Mathis dissenters predicted that this approach would prove a challenge for lower courts because state law is often unclear about whether statutory language lists alternative elements or alternative means.

This Note surveys the court of appeals cases that applied Mathis in the first fifteen months after it was decided. It finds that the Mathis rule has been more workable than the dissenters predicted, but also not straightforward to apply in all cases. The Note then identifies three aspects of prior-conviction doctrine that remain unsettled after Mathis and proposes solutions. First, courts should not feel obligated to resolve the means–element question when the record of prior conviction would not support an enhanced sanction under either interpretation. Second, courts should be permitted to impose prior-conviction enhancements even in the presence of some legal uncertainty about the nature of the prior conviction, so long as there is no factual uncertainty about what the defendant was convicted of. And third, facts alleged in indictments alone are never sufficient to establish factual certainty about the nature of a prior offense.


In December 1980, an Iowa prosecutor charged Richard Mathis with second-degree burglary, “committed as follows: . . . These men broke into an occupied structure and a place where something of value is kept, to-wit: the house and garage of Allen Harvey . . . .” 1 Joint Appendix at 60, Mathis v. United States, 136 S. Ct. 2243 (2016) (No. 15-6092) [hereinafter Mathis Joint Appendix], 2016 WL 4524465. Mathis then pleaded guilty to second-degree burglary and was incarcerated. 2 Id. at 61. This decades-old conviction took on new significance in 2014, when Mathis pleaded guilty to a federal charge of being a felon in possession of a firearm. 3 Id. at 44, 58. The sen­tencing judge had to determine whether Mathis’s prior convictions made him an “armed career criminal” and therefore subject to a fifteen-year mandatory minimum penalty. 4 Id. at 23–24.

The key point of contention was whether the prior convictions were for a burglary of a “building,” which is required for the prior convictions to trigger the mandatory minimum in the Armed Career Criminal Act (ACCA). 5 See Mathis, 136 S. Ct. at 2250–51 (describing the lower courts’ opinions and the question presented). At first the answer seems obvious: Mathis pleaded guilty to burgling the house of Allen Harvey. Houses are buildings, and therefore the prior conviction should qualify. But that is not quite right. Mathis did not plead guilty to burgling Allen Harvey’s house or to any other specific facts; rather, the district court accepted his plea of guilty simply “to the charge of burglary in the 2nd degree.” 6 Mathis Joint Appendix, supra note 1, at 61.

When statutes attach downstream consequences to a certain type of conviction, how should courts determine whether a defendant’s prior conviction falls within the statutory definition? That question has persis­tently vexed courts asked to apply such statutes. 7 See, e.g., Taylor v. United States, 495 U.S. 575, 580 (1990) (describing the challenge of determining what qualifies as a prior conviction for “burglary”); see also infra Part I (tracing the development of federal courts’ approaches to prior convictions). And the answer has enormous human consequences. ACCA provides a much-litigated example: While a first-time felon-in-possession is subject to a ten-year maximum sentence, 8 18 U.S.C. § 924(a)(2) (2012). a defendant given the “career criminal” label faces a fifteen-year minimum sentence. 9 Id. § 924(e). Simi­larly, courts must determine the nature of prior convictions in order to apply the federal sentencing guidelines’ various criminal-history enhancements. 10 See, e.g., U.S. Sentencing Guidelines Manual § 2K2.1(a) (U.S. Sentencing Comm’n 2016) (increasing the applicable sentencing guidelines range if the defendant has “felony convictions of either a crime of violence or a controlled substance offense”); id. § 4B1.1 (describing the “career offender” enhancement). And deportation statutes require courts to consider whether a potential deportee has been convicted of an “aggravated felo­ny” or a “crime involving moral turpitude.” 11 See 8 U.S.C. § 1229b(a) (2012) (prohibiting the Department of Justice from canceling removal of a permanent resident who has been convicted of an aggravated felony); id. § 1229b(b) (prohibiting the Department of Justice from canceling removal of a nonpermanent resident who has been convicted of a crime involving moral turpitude).

In each of these areas, a statute defines the types of offenses that qualify as predicates for a legal consequence. Federal judges must then examine records of prior convictions—records that reflect the varied criminal law and practice of all fifty states 12 See President’s Comm’n on Law Enf’t & Admin. of Justice, The Challenge of Crime in a Free Society 7 (1967), [http://] (“Every village, town, county, city, and State has its own criminal justice system, and there is a Federal one as well. All of them operate somewhat alike. No two of them operate precisely alike.”).
—and determine whether those prior convictions fit the statutory definition. 13 Alina Das, The Immigration Penalties of Criminal Convictions: Resurrecting Categorical Analysis in Immigration Law, 86 N.Y.U. L. Rev. 1669, 1675–76 (2011) (describing such categorization in immigration cases); David C. Holman, Violent Crimes and Known Associates: The Residual Clause of the Armed Career Criminal Act, 43 Conn. L. Rev. 209, 218 (2010) (describing such sentencing in ACCA cases); Doug Keller, Why the Prior Conviction Sentencing Enhancements in Illegal Re-Entry Cases Are Unjust and Unjustified (and Unreasonable Too), 51 B.C. L. Rev. 719, 760–61 (2010) (describing such sentencing under the federal sentencing guidelines). The manner in which judges approach this task determines whether individuals are sub­ject to serious consequences such as deportation or increased terms of incarceration.

This Note examines the Supreme Court’s latest attempt to provide a framework for evaluating prior convictions, Mathis v. United States. In short, Mathis made clear that judges determining whether a prior convic­tion triggers a statutory consequence may examine only the legal ele­ments of the prior conviction, not the factual means of commission. 14 Mathis v. United States, 136 S. Ct. 2243, 2257 (2016) (“ACCA involves, and involves only, comparing elements. Courts must ask whether the crime of conviction is the same as, or narrower than, the relevant generic offense. They may not ask whether the defendant’s conduct—his particular means of committing the crime—falls within the generic definition.”). The dissenters argued that this approach would be difficult to apply because state law is often unclear on which facts are “elements” and which are merely “means.” 15 Id. at 2263 (Breyer, J., dissenting) (“The majority’s approach, I fear, is not practical.”); id. at 2268–69 (Alito, J., dissenting) (arguing it will be difficult for federal judges to make the means–element distinction because “state-court cases on the question are rare”). This Note adds four contributions to the debate. First, it finds that Mathis has proven workable in early applica­tions of its approach, though it does often require federal judges to interpret ambiguous state law. 16 See infra section II.A. Second, it argues that the judicial bur­den of applying the Mathis rule can be lessened by rejecting dicta in the majority opinion that would treat this inquiry into state law as a “threshold” issue that must be resolved before examining the record documents of the defendant’s prior conviction. 17 See infra section III.A. Third, it argues that courts should be permitted to impose prior-conviction enhancements even in the presence of some legal uncertainty about the nature of the prior conviction, so long as there is no factual uncertainty about whether the defendant admitted or was convicted of conduct that would constitute a qualifying offense. 18 See infra section III.B. And fourth, this Note argues that facts alleged in indictments alone are never suffi­cient to establish factual certainty about the nature of a prior offense. 19 See infra section III.C.

Part I summarizes the Court’s long-evolving approach to classifica­tion of prior convictions. Part II evaluates the court of appeals cases that applied Mathis in the six months after it was decided, reaching the find­ing that the majority’s rule has proven workable, though not always straightforward in application. Part III presents recommended interpre­tive clarifications of Mathis.

I. The Challenge of Recidivist Sentencing in a Federal System

Recidivist sentencing rules are a familiar feature of the American criminal justice system, most famously exemplified by “three-strikes” laws. 20 See 18 U.S.C. § 924(e)(2) (2012) (imposing enhanced penalties for felon-in-possession offenders with three qualifying prior convictions); Erwin Chemerinsky, Cruel and Unusual: The Story of Leandro Andrade, 52 Drake L. Rev. 1, 4–5 (2003) (noting that “[e]very state has some form of recidivist sentencing law” and that during the 1990s many states enacted more punitive statutes termed “‘three strikes and you’re out’” laws); Julian V. Roberts, The Role of Criminal Record in the Sentencing Process, 22 Crime & Just. 303, 310 (1997) (describing state “statutes that provide for enhanced sentences for offenders with prior convictions”). Such sentencing rules impose longer sentences on offenders with a pattern of criminal behavior, reasoning that recidivists are more culpa­ble 21 See Youngjae Lee, Repeat Offenders and the Question of Desert, in Previous Convictions at Sentencing 49, 49–51 (Julian V. Roberts & Andrew von Hirsch eds., 2010) (noting the federal sentencing guidelines partly justify their criminal-history enhancements on retributivist grounds and articulating a retributivist justification for such enhancements). and that incarcerating habitual offenders is an effective way to decrease crime. 22 See Roberts, supra note 20, at 316–17 (describing the utilitarian justification that “previous criminal conduct is predictive of future offending”). These rules raise a host of challenging policy ques­tions. 23 Consider four such questions: (1) What sorts of past crimes predict future offending? See Megan C. Kurlychek, Robert Brame & Shawn D. Bushway, Scarlet Letters and Recidivism: Does an Old Criminal Record Predict Future Offending?, 5 Criminology & Pub. Pol’y 483, 483 (2006) (finding the predictive impact of prior convictions fades after about seven years); Michael S. Vigorita, Prior Offense Type and the Probability of Incarceration, 17 J. Contemp. Crim. Just. 167, 167–69, 186 (2001) (finding, in a sample of state cases, no effect of prior-offense similarity on the likelihood of discretionary incarceration). (2) How many offenses are necessary to indicate that a defendant is a habitual offender? See Davis Weisburd, Elin Waring & Ellen F. Chayet, White-Collar Crime and Criminal Careers 52–57 (2001) (finding marked differences in the demographic characteristics of “chronic offenders” with three or more arrests and “low-frequency offenders” with only one or two). (3) Are prior convictions a morally or legally improper basis for enhancing sentences? See Evan Tsen Lee, Mathis v. U.S. and the Future of the Categorical Approach, 101 Minn. L. Rev. Headnotes 263, 274–77 (2016) [hereinafter Lee, Future of the Categorical Approach], [] (suggesting that recidivist enhancements are multiple punishments for the same crime, raising double jeopardy concerns); Claudio Tamburrini, What’s Wrong with Recidivist Punishments?, in Recidivist Punishments 63, 74–75 (Claudio Tamburrini & Jesper Ryberg eds., 2012) (rejecting classic arguments in favor of recidivist enhancements). (4) How large should recidivist enhancements be? See Amy Baron-Evans et al., Deconstructing the Career Offender Guideline, 2 Charlotte L. Rev. 39, 49 (2010) (arguing the career-offender guideline is “more severe than necessary to achieve the purposes of sentencing”). This Note deals with a question of implementation that comes after judgments about which, if any, prior convictions should carry down­stream consequences. That is, how does a sentencing judge know that a defendant has a qualifying prior conviction? After all, in a federal system, a defendant may have prior convictions under a variety of different state or federal criminal statutes. Many of these statutes seek to prevent similar conduct—for example, assault or robbery—but define violations of the law in slightly different ways.

There are three key contexts in which federal prior-offense categori­zation is repeatedly litigated: (1) sentencing under ACCA, (2) sentencing under the federal sentencing guidelines, and (3) cancellation of removal proceedings under 8 U.S.C. § 1229b. 24 See supra notes 8–11 (citing these provisions); infra Part II (detailing recent cases in which federal courts have confronted the issue of prior-offense categorization). In each context, the relevant text provides a different defini­tion of the qualifying predicate offenses. 25 See, e.g., 8 U.S.C. § 1229b(a)(3) (2012) (“aggravated felony”); 18 U.S.C. § 924(e) (2012) (“violent felony”); U.S. Sentencing Guidelines Manual § 4B1.1(a) (U.S. Sentencing Comm’n 2016) (“crime of violence or a controlled substance offense”). Similarly, 8 U.S.C. § 1229b(b)(1)(C) incorporates the definition of a “crime involving moral turpitude” from 8 U.S.C. §§ 1182(a), 1227(a)(2). But the statutes and regula­tions leave unaddressed the important question of how sentencing judges should determine whether court records adequately establish that a defendant was convicted of a qualifying prior offense. This Part intro­duces the judicially created doctrine that has developed to enable such prior-offense categorization. Interestingly, courts employ the same ap­proach in all three contexts 26 See Mathis v. United States, 136 S. Ct. 2243, 2248 (2016) (applying the categorical approach in the ACCA context); Gonzales v. Duenas-Alvarez, 549 U.S. 183, 185–86 (2007) (applying the categorical approach in the immigration context); United States v. Giggey, 551 F.3d 27, 38–39 (1st Cir. 2008) (noting “circuits uniformly apply a categorical approach” in determining whether prior convictions trigger provisions of the federal sentencing guidelines). But see Nijhawan v. Holder, 557 U.S. 29, 36 (2009) (concluding, based on statutory differences, that certain aspects of an immigration statute require “circumstance-specific” rather than categorical treatment (internal quotation marks omitted)). even though the Sixth Amendment does not constrain judicial fact-finding under advisory sentencing guidelines 27 See Mathis, 136 S. Ct. at 2252 (noting the Sixth Amendment prohibition on judicial fact-finding that increases the maximum possible sentence (citing Apprendi v. New Jersey, 530 U.S. 466, 490 (2000))); United States v. Booker, 543 U.S. 220, 246 (2005) (Breyer, J., delivering the opinion of the Court in part) (applying the Sixth Amendment to render federal sentencing guidelines advisory). and does not apply at all in immigration proceedings. 28 See Ali v. Mukasey, 521 F.3d 737, 741 (7th Cir. 2008) (reasoning that because immigration proceedings “are not criminal prosecutions,” Apprendi does not apply); Jennifer Lee Koh, The Whole Better than the Sum: A Case for the Categorical Approach to Determining the Immigration Consequences of Crime, 26 Geo. Immigr. L.J. 257, 276 (2012) (noting that in the immigration context, “the Sixth Amendment and other constitutional criminal protections do not apply,” though characterizing the doctrine that deportation proceedings are noncriminal as “deeply criticized”).

Section I.A briefly describes two sentencing archetypes, conduct sen­tencing and conviction sentencing, to illuminate the policy trade-offs in­volved in charting a middle course. Section I.B describes the limits that Sixth Amendment doctrine imposes on sentencing policy. These sections provide the practical and constitutional background that has shaped the Court’s doctrine of prior-conviction sentencing. Section I.C lays out the pre-Mathis evolution of that doctrine. It traces first the “categorical ap­proach,” in which courts determine only whether the statute defining a defendant’s prior conviction necessarily (that is, “categorically”) meets the definition that triggers a later penalty. It then turns to the “modified cat­egorical approach,” in which courts can examine certain documents from the prior conviction, using them to determine which subpart of a statute the defendant was convicted under. Finally, section I.D explores how courts determine when to use the pure categorical approach and when to use the modified categorical approach. Section I.D also exam­ines the five opinions written in Mathis, which made various predictions about how the majority’s rule would work in practice.

A. Prior Conduct and Prior Convictions

It is helpful at the outset to imagine two possible recidivist sentenc­ing regimes: conduct sentencing and conviction sentencing. 29 This section is substantially similar to Justice Breyer’s description of “pure ‘real offense’” and “pure ‘charge offense’” sentencing in Mathis, 136 S. Ct. at 2263 (Breyer, J., dissenting). Under a pure conduct-sentencing regime, the judge would impose a sentence based not only on the instant offense but also on prior-offense conduct. The practical downside of such a regime is obvious: It would entangle the parties in mini-trials over past conduct. Alternatively, imagine a pure conviction-sentencing regime. Under this system, the federal sentencing judge would not hear any evidence about prior convictions but would look only to the judgments entered in the previous cases. That is, the federal judge would not inquire into the real-world facts of what hap­pened in the prior cases but would consider only the legal facts of the defendant’s prior convictions. If the prior convictions qualified as prior offenses under the recidivist sentencing provision, then the judge would enhance the sentence accordingly.

B. The Sixth Amendment’s Limits on Judicial Fact-Finding: Apprendi and Almendarez-Torres

The Sixth Amendment serves as a critical backdrop to recidivist sen­tencing. In Apprendi v. New Jersey, the Court established that “[o]ther than the fact of a prior conviction, any fact that increases the penalty for a crime beyond the prescribed statutory maximum must be submitted to a jury, and proved beyond a reasonable doubt.” 30 530 U.S. 466, 490 (2000). In Apprendi, the de­fendant had pleaded guilty to a firearm charge that carried a penalty of five to ten years, but then the sentencing judge independently found that the defendant had committed the crime with a purpose to intimidate a racial group and imposed a twelve-year sentence. 31 Id. at 468–71. The Court described this as “an unacceptable departure from the jury tradition” and found it in violation of the Sixth and Fourteenth Amendments. 32 Id. at 497.

As the first quote in the preceding paragraph indicates, there is an exception for prior convictions. In an earlier case, Almendarez-Torres v. United States, the Court upheld a sentence that the judge had imposed after independently finding that the defendant had a prior conviction for an “aggravated felony.” 33 523 U.S. 224, 227 (1998). The baseline statute authorized a two-year sentence for illegal reentry after deportation but allowed a sentence of up to twenty years “if the initial ‘deportation was subsequent to a convic­tion for commission of an aggravated felony.’” 34 Id. at 226 (quoting 8 U.S.C. § 1326(b)(2) (1994)). In its reasoning, the Court explained that recidivism “is a traditional, if not the most tradi­tional, basis for a sentencing court’s increasing an offender’s sentence.” 35 Id. at 243.

C. The Evolution of the Modified Categorical Approach

1. The Categorical Approach. — The Supreme Court began articulating its modern doctrine of prior-offense categorization in Taylor v. United States. 36 See 495 U.S. 575, 600 (1990). There the Court held that federal courts should use a model that mostly embodies a conviction-sentencing regime. 37 See id. Specifical­ly, the Court explained that ACCA “mandates a formal categorical ap­proach, looking only to the statutory definitions of the prior offenses, and not to the particular facts underlying those convictions.” 38 Id.

The Court reached this conclusion for three reasons. First, the stat­ute refers to previous “convictions,” not prior conduct that would consti­tute a violation of the law. 39 Id. Second, the legislative history did not indicate that Congress contemplated the sort of in-depth fact-finding that would be necessary to examine the details of a defendant’s prior conduct. 40 Id. at 601. Third, the Court explained that “the practical difficulties and potential unfairness of a factual approach are daunting.” 41 Id. Though Taylor predates Apprendi, the Court also suggested that judicial imposi­tion of a longer sentence based on independent review of the facts might abridge the right to a jury trial. 42 Id.

The Court noted that the defendant’s prior convictions might have occurred under statutes that were “narrower” or “broad[er]” than the language that Congress used to define a qualifying prior conviction. 43 Id. at 599. A narrower definition is one that regulates conduct that necessarily fits within the federal definition. A broader definition is one that might have produced a conviction for conduct that Congress did not mean to in­clude as a qualifying prior offense. In the latter case, the Court said that the categorical approach “may permit the sentencing court to go beyond the mere fact of conviction in a narrow range of cases where a jury was actually required to find all the elements of [the] generic [offense].” 44 Id. at 602. For example, the Court explained of the burglary statute at issue in Taylor:

[I]n a State whose burglary statutes include entry of an auto­mobile as well as a building, if the indictment or information and jury instructions show that the defendant was charged only with a burglary of a building, and that the jury necessarily had to find an entry of a building to convict, then the Government should be allowed to use the conviction for enhancement. 45 Id.

This investigation into the nature of a prior conviction was later termed the “modified categorical approach.” 46 See, e.g., Tokatly v. Ashcroft, 371 F.3d 613, 620 (9th Cir. 2004).

2. The Modified Categorical Approach. — The Court’s endorsement of a limited factual inquiry raised questions of how precisely judges should conduct such inquiry, an issue addressed in Shepard v. United States. 47 544 U.S. 13, 16 (2005). The government had alleged that Reginald Shepard was a felon in possession of a firearm and subject to the ACCA enhancement due to four prior burglary convictions. 48 Id. The parties agreed that the Massachusetts statute, under which Shepard had been previously convicted, punished a broad­er range of conduct than the ACCA predicates. 49 Id. at 17. But the government argued that the convictions were necessarily for qualifying conduct un­der the modified categorical approach. 50 Id. at 16–17. Because Shepard had pleaded guilty to the prior burglaries, there were no jury instructions available, and the government asked the district court to consider police reports as evidence that Shepard’s prior convictions should qualify. 51 Id.

The Court rejected the use of police reports, reasoning that it would open the door to the broader factual inquiries that Taylor meant to pre­vent. 52 Id. at 22–23. Ultimately, the Court explained that the evidence judges could consider in applying the modified categorical approach to guilty pleas was “limited to the terms of the charging document, the terms of a plea agreement or transcript of colloquy between judge and defendant in which the factual basis for the plea was confirmed by the defendant, or to some comparable judicial record of this information.” 53 Id. at 26.

D. Alternative Elements and Alternative Means

Though Shepard provided clear guidance on what documents courts could consider in applying the modified categorical approach, a further circuit split developed concerning when courts could reach the modified categorical approach at all. 54 See Descamps v. United States, 133 S. Ct. 2276, 2283 & n.1 (2013) (noting the presence of this circuit split). Some courts held that they could employ the modified categorical approach whenever a statute criminalized a broader range of conduct than the federal generic offense. 55 See, e.g., United States v. Armstead, 467 F.3d 943, 948 (6th Cir. 2006), abrogated by Descamps, 133 S. Ct. 2276. Others held that when a statute was merely broad but did not set out a disjunctive list of alternative forms of violation, only the pure categorical approach was permissible. 56 See, e.g., United States v. Beardsley, 691 F.3d 252, 268–69 (2d Cir. 2012).

This circuit split rested on the difference between statutes that may simply be violated in various ways and those that are textually divisible. Conceptually, every criminal statute may be violated in a number of dif­ferent ways. Take, for example, New York’s third-degree assault statute, which provides that a defendant is guilty of the offense when “[w]ith in­tent to cause physical injury to another person, he causes such injury to such person or to a third person.” 57 N.Y. Penal Law § 120.00(1) (McKinney 2016). A defendant might violate this stat­ute by intentionally punching a victim or by intentionally shooting a vic­tim; either would be sufficient for a conviction. As a result, for a later judge required to determine whether the defendant had a prior conviction involving the use of a firearm, 58 Cf. 18 U.S.C. § 924(e)(2)(B) (2012) (defining an act of juvenile delinquency as a “violent felony” only if it involves “the use or carrying of a firearm, knife, or destructive device”). the bare fact of conviction would not reveal whether a New York third-degree assault conviction qualifies. Compare New York’s second-degree assault statute, which pro­vides that a defendant violates the statute when “[h]e recklessly causes physical injury to another person who is a child under the age of eighteen by intentional discharge of a firearm, rifle or shotgun.” 59 N.Y. Penal Law § 120.05(4–a). The second-degree assault statute also provides numerous other ways in which it can be violated, but the presence of particular text specifying the in­volvement of a firearm makes it possible that the record of conviction will reveal whether the defendant’s prior offense involved the use of a firearm.

In two recent cases, Descamps v. United States and Mathis v. United States, the Supreme Court decided first that only textually divisible stat­utes allow a sentencing judge to apply the modified categorical ap­proach 60 Descamps, 133 S. Ct. at 2293. and, second, that the different textual components must be not “means” but “elements.” 61 Mathis v. United States, 136 S. Ct. 2243, 2247–48 (2016).

1. Requiring a Textual Basis for Divisibility—Descamps v. United States. — In Descamps, the Supreme Court resolved the circuit split out­lined above, holding that the modified categorical approach was permis­sible only when a statute is “divisible.” 62 Descamps, 133 S. Ct. at 2293. The Court explained that a divisible statute “sets out one or more elements of the offense in the alternative—for example, stating that burglary involves entry into a build­ing or an automobile.” 63 Id. at 2281. The Court acknowledged that indivisible statutes also implicitly set out a list of means of commission; 64 Id. at 2289. thus, the only con­ceptual difference between indivisible statutes and divisible ones is that the divisible statute’s alternatives are explicit. 65 Id. A further example may help illuminate the distinction. Suppose that a court must determine whether a prior conviction “is burglary.” Cf. 18 U.S.C. § 924(e)(2)(B) (2012) (defining “violent felony” to include any crime that “is burglary”). Suppose further that the defendant has a prior conviction for theft under Texas law, which means that he “unlawfully appropriate[d] property with intent to deprive the owner of property.” Tex. Penal Code § 31.03(a) (2015). This statute can be violated by means that resemble burglary. See Rice v. State, 861 S.W.2d 925, 925 (Tex. Crim. App. 1993) (affirming simultaneous convictions for theft and burglary arising out of “the same transaction”). But the theft statute can also be violated in ways that would not satisfy the elements of burglary. Compare Tex. Penal Code § 30.02(a) (requiring, for a burglary conviction, entry to or remaining in a habitation or building), with Moron v. State, 779 S.W.2d 399, 401 (Tex. Crim. App. 1985) (affirming a theft conviction with no evidence that the defendant entered a building). Because there is no text in the theft statute that would allow a finder of fact to specify that it was violated by burglary-like means, the theft statute would be indivisible, such that the sentencing judge in a subsequent case could not employ the modified categorical approach to determine whether the prior conviction qualified as “burglary.” But the Court reasoned that “only divisible statutes enable a sentencing court to conclude that a jury (or judge at a plea hearing) has convicted the defendant of every element of the generic crime.” 66 Descamps, 133 S. Ct. at 2290. Indeed, the Court said, allowing the modified categorical approach for indivisible statutes “would altogether collapse the distinction between a categorical and a fact-specific ap­proach” because the sentencing judge could always “imaginatively trans­form[]” a broader statute to cover a range of narrow factual alterna­tives. 67 Id. at 2290–91.

Justice Alito, dissenting, argued that it would often be difficult to de­termine whether a statute was divisible. 68 Id. at 2301 (Alito, J., dissenting). He observed that statutes often list alternatives even though the prosecution need not prove any particular one of these alternatives beyond a reasonable doubt. 69 Id. He offered as an example a Michigan statute that “criminalizes assault with ‘a gun, revolver, pistol, knife, iron bar, club, brass knuckles, or other dangerous weapon’” 70     [1]. Id. (quoting Mich. Comp. Laws Ann. § 750.82(1) (West 2004)). and noted that, despite this seemingly divisible list, Michigan courts have held that the prosecution merely needs to prove that a defendant used a dangerous weapon, not any particular weapon. 71 Id. (citing People v. Avant, 597 N.W.2d 864, 869 (Mich. Ct. App. 1999)). In a subsequent prosecution in which it was important whether the prior conviction involved use of a firearm, for example, it would be critical whether the statute listed alternative elements or alternative means. Con­sequently, Justice Alito argued, the Descamps majority would complicate lower courts’ application of ACCA because “[t]he only way to be sure whether particular items are alternative elements or simply alternative means of satisfying an element may be to find cases concerning the cor­rectness of jury instructions that treat the items one way or the other. And such cases may not arise frequently.” 72 Id. at 2301–02.

Responding in a footnote, the majority wrote that it saw “no real-world reason to worry” about the difficulty of distinguishing between means and elements. 73 Id. at 2285 n.2 (majority opinion). The Court elaborated,

Whatever a statute lists (whether elements or means), the doc­uments we approved in Taylor and Shepardi.e., indictment, jury instructions, plea colloquy, and plea agreement—would reflect the crime’s elements. So a court need not parse state law in the way the dissent suggests: When a state law is drafted in the alternative, the court merely resorts to the approved docu­ments and compares the elements revealed there to those of the generic offense. 74 Id.

The text of this footnote, referring to both elements and means, produced a further circuit split: Can a sentencing judge apply the modi­fied categorical approach only when a statute lists alternative elements, or is a statute that lists alternative means sufficient, so long as they are re­flected in the Shepard documents? 75 Mathis v. United States, 136 S. Ct. 2243, 2251 (2016).

2. Requiring Alternative Elements, Not Just Means—Mathis v. United States. — In Mathis, the Court resolved this circuit split by holding that “ACCA disregards the means by which the defendant committed his crime, and looks only to that offense’s elements.” 76 Id. at 2256. As a theoretical mat­ter, it is straightforward to explain the difference between elements and means. As the Mathis Court put it:

“Elements” are the “constituent parts” of a crime’s legal definition—the things the “prosecution must prove to sustain a conviction.” At a trial, they are what the jury must find beyond a reasonable doubt to convict the defendant; and at a plea hear­ing, they are what the defendant necessarily admits when he pleads guilty. Facts, by contrast, are mere real-world things—extraneous to the crime’s legal requirements. . . . They are “circumstance[s]” or “event[s]” having no “legal effect or consequence”: In particular, they need neither be found by a jury nor admitted by a defendant. 77 Id. at 2248 (citations omitted) (quoting Black’s Law Dictionary 634, 709 (10th ed. 2014)).

But as a practical matter, determining whether a given statute lists alternative elements or alternative means is more challenging. The ques­tion is one of state law, and one that most state courts approach as a question of legislative intent: That is, did the legislature intend to re­quire proof beyond a reasonable doubt on that specific fact (in which case it is an element) or not (in which case it is merely a means)? 78 See, e.g., State v. Peterson, 230 P.3d 588, 591 (Wash. 2010) (“[T]here simply is no bright-line rule by which the courts can determine whether the legislature intended to provide alternate means of committing a particular crime. Instead, each case must be evaluated on its own merits.” (internal quotation marks omitted) (quoting State v. Klimes, 73 P.3d 416, 422 (Wash. Ct. App. 2003))). Legislatures often simply draft statutes in the alternative, without clearly indicating whether the alternatives are elements or means. 79 Jessica A. Roth, Alternative Elements, 59 UCLA L. Rev. 170, 180–83 (2011) (noting this trend in drafting and that “courts charged with deciding whether a particular statute or series of statutes create one or more than one offense will frequently reach different conclusions”).

The Mathis majority provided three sources of state law that courts could look to in making the means–element distinction: state-court deci­sions, the text of the statute, and the record of the prior conviction. 80 Mathis, 136 S. Ct. at 2256–57. This Note will refer to these as the “Mathis tools,” and the core project of Part II is determining how well these tools have served lower courts tasked with distinguishing statutory elements from means.

3. Conflicting Predictions About the Workability of Mathis. — Mathis produced five separate opinions: (1) Justice Kagan’s majority opinion, signed by five members of the Court, (2) Justice Kennedy’s concurrence, (3) Justice Thomas’s concurrence, (4) Justice Breyer’s dissent, joined by Justice Ginsburg, and (5) Justice Alito’s dissent. The majority and con­curring opinions reveal important differences about the justification for requiring alternative elements. And the dissenting opinions propose al­ternative methods of applying ACCA that they think are more workable and consistent with congressional purpose.

Justice Kagan offered three key rationales for limiting the modified categorical approach to elements: (1) ACCA’s text refers to “convictions” rather than conduct, (2) judicial fact-finding would raise Sixth Amendment trial-by-jury concerns, and (3) non-elemental facts of prior convictions are unreliable because the defendant had little incentive to correct them. 81 Id. at 2252–53.

Justice Kennedy’s concurrence rejected the second of these ration­ales, noting that he believes Apprendi was wrongly decided and “does not compel the elements based approach.” 82 Id. at 2258 (Kennedy, J., concurring). Justice Thomas, by contrast, explained that he believes the Almendarez-Torres exception to Apprendi is wrong, and thus the Sixth Amendment prohibits any form of judicial reliance on prior convictions. 83 Id. at 2258–59 (Thomas, J., concurring). He thus joined the majority approach because it at least limits judges’ ability to make such factual determina­tions. 84 Id. at 2259.

Justice Breyer rejected all three of the majority’s rationales. In his view, ACCA’s reliance on “convictions” was designed as a “practical” so­lution to allow sentencing judges to quickly determine whether prior convictions involved the sort of behavior that Congress wanted to tar­get. 85 Id. at 2263 (Breyer, J., dissenting). Justice Breyer therefore attacked the majority’s approach primarily on the ground of workability, predicting that if courts must research whether a pro­vision lists elements or means under state law, “[w]hat was once a simple matter will produce a time-consuming legal tangle.” 86 Id. at 2264. He also rejected the Apprendi rationale, reasoning that a fact must be proven beyond a reasonable doubt if it was the only alternative that the prior prosecution charged. 87 Id. at 2265. Perhaps most controversially, Justice Breyer considered the charging document’s allegation of a qualifying offense sufficient indica­tion that the defendant’s guilty plea necessarily produced a conviction for a qualifying offense. 88 Id. at 2260 (“[T]he federal sentencing judge can look at the charging documents (or plea colloquy) to see whether ‘the defendant was charged only with a burglary of a building.’” (quoting Taylor v. United States, 495 U.S. 575, 602 (1990))).

Justice Alito, having attacked the majority’s baseline rationales in Descamps, 89 Descamps v. United States, 133 S. Ct. 2276, 2295–301 (2013) (Alito, J., dissenting). advanced two points at greater length in Mathis. First, to the majority’s assertion that determining whether a statute listed alternative elements or alternative means would often be easy, Justice Alito bluntly replied, “Really?” 90 Mathis, 136 S. Ct. at 2269 (Alito, J., dissenting). By contrast, Justice Alito argued that the majority’s approach would require lower courts to delve into an area of state law that is notoriously uncertain. 91 Id. at 2269 n.3. Further, he argued that the Court’s ap­proach produced “results that Congress could not have intended.” 92 Id. at 2268. Specifically, Justice Alito suggested that defendants who engaged in the same conduct would be treated differently depending on the drafting of the state statute under which they were convicted. Justice Alito character­ized this result as the ultimate example of “pointless formalism.” 93 Id. at 2271.

The differences between these five opinions motivate the questions that Part II explores. First, has the Mathis approach proved workable for lower courts to apply? And second, how have courts operationalized the three interpretive tools that Mathis suggests?

II. Mathis in Practice

This Part examines the court of appeals cases that have applied the Mathis framework to determine whether a statute is divisible and reaches three conclusions. First, contrary to the dissenters’ workability concerns, applying Mathis is within the core competencies of the federal judiciary, and the interpretive tools that it provides produce predictable results in most cases. Second, though the Mathis inquiry requires parsing of state law to determine whether a statute is divisible, a handful of court of ap­peals cases omit such analysis and make only a conclusory statement on the question of divisibility. Third, questions remain about how to apply portions of the Mathis opinion regarding the level of certainty required about the nature of the prior conviction and the use of record docu­ments.

A. Cases Illustrating the Workability of Mathis

1. Defining Workability. — What precisely does it mean for a standard to be workable? 94 The word “workable” appears only in Justice Breyer’s Mathis dissent as a part of a quotation from Fourth Circuit Judge Paul Niemeyer. See id. at 2264 (Breyer, J., dissenting) (stating there are no “clear and workable standards” for applying the means–element distinction (quoting Omargharib v. Holder, 775 F.3d 192, 200 (4th Cir. 2014) (Niemeyer, J., concurring))). This Note uses “workable” throughout as a shorthand for the concerns that Justices Breyer and Alito expressed about lower-court judges’ ability to determine whether a statute sets out alternative elements or alternative means. See infra notes 95–97. Justice Breyer criticized the Mathis majority’s approach on the grounds that it would “produce a time-consuming legal tangle.” 95 Mathis, 136 S. Ct. at 2264 (Breyer, J., dissenting). Justice Breyer added that “lower court judges have criticized the approach the majority now adopts” and quoted Judge Niemeyer for the proposition that “[b]ecause of the ever-morphing analysis and the increasingly blurred articulation of applicable standards, [lower courts] are being asked to decide, without clear and workable standards, whether disjunctive phrases in a criminal law define alternative elements of a crime or alternative means of committing it.” Id. (internal quotation marks omitted) (quoting Omargharib, 775 F.3d at 200 (Niemeyer, J., concurring)). In a similar vein, while on the First Circuit, then-Judge Breyer observed that the criminal justice system “must be administratively workable” and that “[t]he more the system recognizes the tendency to treat different cases differently, however, the less manageable the sentencing system becomes.” Stephen Breyer, The Federal Sentencing Guidelines and the Key Compromises upon Which They Rest, 17 Hofstra L. Rev. 1, 13 (1988). Further, Justice Breyer feared that the Mathis approach would prove to be “not practical” in part because “there are very few States where one can find authoritative judicial opinions that decide the means/element question.” 96 Mathis, 136 S. Ct. at 2263–64 (Breyer, J., dissenting). Similarly, Justice Alito expressed concern that “[t]he Court’s approach calls for sentencing judges to delve into pointless abstract questions” and that lower courts would struggle to apply the means–element distinction in the frequent cases that raise the applicability of the modified categorical approach. 97 Id. at 2268 (Alito, J., dissenting).

To evaluate these arguments, one needs an account of what makes a standard “practical” or “workable.” Though none of the opinions in Mathis provide such an account, the Court has directly confronted work­ability in the context of the political question doctrine, invoking the concept of “judicially discoverable and manageable standards.” 98 Vieth v. Jubelirer, 541 U.S. 267, 277 (2004) (quoting Baker v. Carr, 369 U.S. 186, 217 (1963)); see also id. at 288 (concluding a proposed test for identifying impermissible partisan gerrymandering was “not judicially manageable”). Professor Richard Fallon observes that the Court applies this concept in other contexts as well and “sometimes expressly justifies its selection [of a judicial test] as more manageable than the alternatives.” 99 Richard H. Fallon, Jr., Judicially Manageable Standards and Constitutional Meaning, 119 Harv. L. Rev. 1274, 1297–98 (2006); see also Armstrong v. Exceptional Child Ctr., 135 S. Ct. 1378, 1385 (2015) (invoking the “judicially unadministrable nature” of statutory text as a reason for adopting a construction of the statute that would not require judicial enforcement). Though the Mathis majority did not justify its selection of an elements-based approach based on its workability, it did feel the need to say that applying the test would be “easy” in many cases and would only rarely result in “indeterminacy.” See Mathis, 136 S. Ct. at 2256–57. Fallon iden­tifies the following “practical desiderata” as probative on the issue of whether a standard is judicially manageable: (1) whether it has sufficient “analytical bite” to be rigorously applied, (2) whether it can generate predictable and consistent results, and (3) whether it requires the courts to make empirical judgments beyond their competence. 100 Fallon, supra note 99, at 1285–96. This accounting simplifies Fallon’s criteria in two respects. First, his article avers that, to be manageable, a standard must be capable of being understood. That cannot explain the disagreement in Mathis, as the majority and dissents alike treated the means–element distinction as intelligible. See Mathis, 136 S. Ct. at 2250 (explaining the difference between means and elements); id. at 2261 (Breyer, J., dissenting) (defining the difference between means and elements in the context of jury unanimity); id. at 2268 (Alito, J., dissenting) (noting “[t]he distinction between an ‘element’ and a ‘means’ is important in” the context of jury unanimity). Second, the above summary omits one factor relating to remedial formulation, which is not relevant in the context of prior-conviction consequences. In these decisions, the remedy is always straightforward: If the defendant was not necessarily convicted of a predicate offense, simply do not impose the enhanced sanction. Finally, Fallon suggests that courts “ultimately make all-things-considered judgments” about which doctrinal test to adopt after considering the preceding indi­cia of manageability along with the other costs and benefits of the pro­posed rule. 101 Fallon, supra note 99, at 1312–13. The following section employs Fallon’s “practical desiderata” to evaluate whether the Mathis rule has been workable as ap­plied by the lower courts.

2. The Mathis Tools Producing Principled Results. — In most cases, the Mathis tools produce predictable results that rest on courts’ core competencies of statutory and common-law interpretation. Of the first 103 court of appeals decisions confronting the means–element distinction under the Mathis framework, only 12 produced split panels. 102 See infra Appendix; see also Kari Hong, The Absurdity of Crime-Based Deportation, 50 U.C. Davis L. Rev. 2067, 2111–14 (finding a circuit on the side of the split that won out in Mathis was no more likely to have split-panel decisions than a circuit more broadly applying the modified categorical approach). The modified categorical approach jurisprudence, for all its complexity, does provide a relatively rule-bound and predictable framework for decision.

Fallon describes the first practical consideration in the Supreme Court’s workability analysis as requiring that a test provide “criteria suffi­cient to make nonarbitrary distinctions.” 103 Fallon, supra note 99, at 1287. A test is not unworkable merely because it is difficult to apply but only if “it requires distinctions for which conceptual resources are lacking in too many instances.” 104 Id. As a paradigmatic example of a test lacking sufficient analytical bite, Fallon points to the Supreme Court’s now-discarded distinction between activities that directly affect interstate commerce and those that affect it only indirectly. Id.; see also United States v. Lopez, 514 U.S. 549, 555–56 (1995) (describing the Court’s abandonment of the direct–indirect distinction). This section finds that, in most cases, the Mathis tools—state cases, statutory text, and the record of conviction—have provided sufficient criteria to make nonarbitrary distinctions in court of appeals cases decided since Mathis.

First, state cases have provided guidance in about sixty percent of the court of appeals cases applying the Mathis framework. 105 See infra Appendix. These cases do not always offer explicit holdings about what facts a jury must find unanimously. Rather, the state cases have arisen in a wide variety of legal contexts. See, e.g., United States v. Ocampo–Estrada, 873 F.3d 661, 668 (9th Cir. 2017) (citing In re Adams, 536 P.2d 473, 479 (Cal. 1975)) (relying on a state case upholding multiple sentences under a single statute); Marinelarena v. Sessions, 869 F.3d 780, 786 (9th Cir. 2017) (quoting People v. Horn, 524 P.2d 1300, 1304 (Cal. 1974)) (relying on a state case requiring juries in conspiracy cases to be instructed on the elements of the alleged objects of the conspiracy); United States v. McMillan, 863 F.3d 1053, 1057 (8th Cir. 2017) (quoting State v. Witherspoon, No. A12–1247, 2013 WL 3284272, at *2 (Minn. Ct. App. July 1, 2013)) (relying on a state sufficiency-of-evidence case that described the elements of the offense); United States v. Hudson, 851 F.3d 807, 809 (8th Cir. 2017) (citing Yates v. State, 158 S.W.3d 798, 801–02 (Mo. Ct. App. 2005)) (relying on a state double jeopardy case that described the elements of the offense). It is important to un­derstand the limitations of this quantitative finding and those that follow. Court of appeals cases do not directly capture how the Mathis tools func­tion in the district courts, 106 See Pauline T. Kim et al., How Should We Study District Judge Decision-Making?, 29 Wash. U. J.L. & Pol’y 83, 84–94 (2009) (observing that quantitative studies of trial courts often improperly assume “that judging at the trial court level is fundamentally the same as judging at the appellate level”). which handle the bulk of federal cases. 107 Lee Epstein, William M. Landes & Richard A. Posner, The Behavior of Federal Judges 208 (2013). Similarly, in examining whether Mathis furnishes sufficiently clear crite­ria to resolve most cases, one should keep in mind that courts of appeals are likely to hear cases that are closer or more challenging than the aver­age case. 108 See George L. Priest & Benjamin Klein, The Selection of Disputes for Litigation, 13 J. Legal Stud. 1, 8, 13–17 (1984) (theorizing that parties are most likely to litigate when the dispute is a close case relative to the legal standard); see also Michael Heise, Federal Criminal Appeals: A Brief Empirical Perspective, 93 Marq. L. Rev. 825, 839–42 (2009) (noting selection effects unique to criminal cases, such as the fact that the Double Jeopardy Clause prevents the government from initiating appeals). Despite these limitations, the frequency with which courts of appeals have found state cases to guide their determination of the means–element distinction shows that such declarations of state law are perhaps not as rare as the Mathis dissenters feared. 109 See Mathis v. United States, 136 S. Ct. 2243, 2268–69 (2016) (Alito, J., dissenting).

One example of such a case—an “easy” case in the words of the Mathis majority 110 See id. at 2256 (majority opinion). —is Gomez-Perez v. Lynch. 111 829 F.3d 323 (5th Cir. 2016). Gomez-Perez sought cancel­lation of removal proceedings against him, a request that the immigra­tion judge had denied, reasoning that his misdemeanor assault was a “crime of moral turpitude.” 112 Id. at 325; see also 8 U.S.C. § 1229b(b)(1)(C) (2012) (preventing discretionary cancellation of removal for aliens convicted of crimes involving moral turpitude). Though this case arose in an immigration proceeding, courts apply the same categorical approach to determine whether a prior conviction meets a statutory definition in the current proceeding. 113 United States v. Mayer, 560 F.3d 948, 951 (9th Cir. 2009) (Kozinski, J., dissenting from the denial of reh’g en banc) (“The categorical approach . . . has been applied with remarkable uniformity to many areas of law. It’s used for ACCA; it’s used for immigration; it’s used for the Sentencing Guidelines.”). Gomez-Perez was convicted under a Texas Penal Code provision that defines assault as “intentionally, knowingly, or recklessly cause[ing] bodily injury to another.” 114 Gomez-Perez, 829 F.3d at 325 (citing Tex. Penal Code Ann. § 22.01(a)(1) (West 2015)). Only crimes committed with the mental state of intent qualify as crimes involving moral turpitude that make a nonpermanent resident ineligible for discretionary cancellation of removal. 115 Mathis, 136 S. Ct. at 2253 n.3; see also Gomez-Perez, 829 F.3d at 325 (noting that Gomez-Perez “sought cancellation [of removal proceedings] as a nonpermanent resident”). In a one-paragraph analysis, the Fifth Circuit said this was not a difficult case under Mathis because “Texas law has definitively answered the ‘means or elements’ question.” 116 Gomez-Perez, 829 F.3d at 327–28. And indeed, the Court of Criminal Appeals of Texas has explicitly written:

The legislature was apparently neutral about which of these three mental states accompanied the forbidden conduct because all three culpable mental states are listed together in a single phrase within a single subsection of the statute. There is no indication that the legislature intended for an “intentional” bodily injury assault to be a separate crime from a “knowing” bodily injury assault or that both of those differ from a “reckless” bodily injury assault. All three culpable mental states are strung together in a single phrase within a single subsection of the statute. All result in the same punishment. They are con­ceptually equivalent. 117 Landrian v. State, 268 S.W.3d 532, 537 (Tex. Crim. App. 2008).

Given such a clear interpretation that the different mental states are means, rather than elements, the Mathis test is straightforward for the federal sentencing judge to apply: The statute is indivisible.

Similarly, the D.C. Circuit resolved a case under the career-offender guideline 118 The courts of appeals also utilize the categorical approach articulated in ACCA cases to decide cases under the career-offender guideline. See United States v. Giggey, 551 F.3d 27, 39 (1st Cir. 2008). in part by reference to state law. 119 See United States v. Sheffield, 832 F.3d 296, 314–15 (D.C. Cir. 2016). The district court had sentenced Dante Sheffield as a career offender in part due to a prior conviction for attempted robbery. 120 Id. at 311. The District of Columbia’s robbery statute can be violated by the use of force (“against resistance”), the threat of force (“by putting in fear”), or by means that do not involve the use of violent force against a person (“by sudden or stealthy seizure or snatching”). 121 D.C. Code § 22-2801 (2017). By contrast, the attempted robbery statute does not dis­tinguish these separate forms. 122 See id. § 22-2802. Instead, as the D.C. Court of Appeals has explained, attempted robbery has only three elements:

(1) [T]he defendant committed an act which was reasonably adapted to the commission of the offense of robbery, (2) at the time the act was committed, the defendant acted with the spe­cific intent to commit the offense of robbery, and (3) the act went beyond mere preparation, and carried the project forward to within dangerous proximity of the criminal end to be sought. 123 Sheffield, 832 F.3d at 315 (internal quotation marks omitted) (quoting Robinson v. United States, 608 A.2d 115, 116 (D.C. 1992)).

Thus, D.C. law makes clear that attempted robbery does not have any divisible form in which the jury would necessarily have to find at­tempted use of physical force, and the statute cannot categorically de­scribe a crime of violence. As these two cases illustrate, state-court deci­sions do often provide sufficient information to make a nonarbitrary distinction between divisible and indivisible statutes.

This is not to say that parsing state law is always straightforward. In United States v. Fogg, the Eighth Circuit considered whether a Minnesota drive-by-shooting conviction was a crime of violence under ACCA. 124 836 F.3d 951, 954 (8th Cir. 2016). The Minnesota statute contained two provisions, one specifying punishment for drive-by shootings and the other specifying greater punishment for such shooting that involved “firing at or toward a person.” 125 Minn. Stat. § 609.66, subdiv. 1e (2016). Only the latter form would qualify as an ACCA predicate, which requires “use of physical force against the person of another.” 126 Fogg, 836 F.3d at 954 (internal quotation marks omitted) (quoting 18 U.S.C. § 924(e)(2)(B)(i) (2012)). In Fogg, as in many of the cases considered in this Note, the prisoner’s prior convictions might once have been adjudicated crimes of violence under ACCA’s “residual clause,” which covered felony convictions that “otherwise involve[] conduct that presents a serious potential risk of physical injury to another.” See 18 U.S.C. § 924(e)(2)(B)(ii). In Johnson v. United States, the Supreme Court held that ACCA’s residual clause was unconstitutionally vague. 135 S. Ct. 2551, 2557 (2015). This residual clause once functioned as an alternative avenue for sentencing judges to conclude that prior convictions qualified as predicate offenses, and its invalidation accounts for much of the increased frequency of litigation over whether statutes are divisible in such a way that they could qualify as predicate offenses under the remaining provisions of the statute. See Welch v. United States, 136 S. Ct. 1257, 1262, 1268 (2016) (noting the “many offenders sentenced under the Armed Career Criminal Act before Johnson was decided” and remanding for consideration of whether the prisoner’s prior convictions qualify under the elements clause of ACCA); United States v. Esprit, 841 F.3d 1235, 1237 (11th Cir. 2016) (describing how Johnson and Mathis had changed the analysis of whether the prisoner’s convictions were crimes of violence); King v. United States, 202 F. Supp. 3d 1346, 1349–52 (S.D. Fla. 2016) (tracing the parties’ arguments in response to recent ACCA decisions, including Johnson). The panel majority concluded that the second provision must be an additional element un­der Apprendi because it allowed higher punishment than otherwise per­missible. 127 Fogg, 836 F.3d at 955. In dissent, Judge Myron Bright cited state precedent that the second provision was a separate “sentencing enhancement,” and that a person thus could be convicted under the statute without proof of the second element. 128 Id. at 960–61 (Bright, J., dissenting) (citing State v. Hayes, 826 N.W.2d 799, 806 (Minn. 2013)). The majority in turn responded that this was a “misread[ing]” of the state case and pointed to other Minnesota cases holding that sentencing enhancements must be proven beyond a reason­able doubt, thus rendering them elemental for ACCA purposes. 129 Id. at 955 (majority opinion).

The foregoing discussion illustrates that the import of state law may not always be clear. 130 For another example of a case in which parsing state law was somewhat challenging, see United States v. Tavares, 843 F.3d 1, 17 (1st Cir. 2016) (predicting, based on higher state and federal precedents, that the Massachusetts Supreme Judicial Court would not follow an intermediate appellate court’s decision that certain facts distinguishing forms of assault and battery are nonelemental). But parsing judicial decisions and statutory text is within the core competency of federal courts. 131 See Zivotofsky ex rel. Zivotofsky v. Clinton, 566 U.S. 189, 201 (2012) (characterizing “careful examination of . . . textual, structural, and historical evidence” as “what courts do”); S.D. ex rel. Dickson v. Hood, 391 F.3d 581, 605 (5th Cir. 2004) (“[S]tatutory analysis does not ‘strain judicial competence;’ it is the sort of work in which courts engage every day.” (quoting Blessing v. Freestone, 520 U.S. 329, 341 (1997)); Sims v. CIA, 642 F.2d 562, 572 (D.C. Cir. 1980) (“Construction of statutes is an area of special judicial competence.”). As Fogg illustrates, it is true that Mathis and the Supreme Court’s other recent ACCA cases have required lower courts to examine new and challenging questions about whether given state statutes set out means or elements. 132 See Transcript of Sentencing Hearing at 13–15, United States v. Fogg, No. 0:14-cr-00249 (D. Minn. Sept. 2, 2015) [hereinafter Fogg Transcript of Sentencing Hearing] (characterizing inquiry into whether a fact is elemental as a “difficult issue” and noting that the Supreme Court’s invalidation of the residual clause is “putting pressure on this issue in a way there’s never been pressure on this issue before”). But, at least when state cases are available, construction of those cases provides adequate conceptual resources for courts to decide those new questions in a principled way.

The second Mathis tool, statutory text, provides further guidance for making the means–element distinction in the absence of state-court precedent. Mathis provides two ways in which the text of a statute could reveal whether an alternatively phrased statute sets out elements or means. 133 Mathis actually suggests a third textual indicator, that “a statute may itself identify which things must be charged,” Mathis v. United States, 136 S. Ct. 2243, 2256 (2016), but no court of appeals case has since relied on such a methodology. For that reason, this Note does not examine how this third indicator affects the application of Mathis. Further, the proposition that anything that “must be charged” consequently is an element, id., seems questionable as a general statement of law. In support of the proposition, Mathis cites as an example of such a statute Cal. Penal Code § 952 (2016), which establishes that “[i]n charging theft it shall be sufficient to allege that the defendant unlawfully took the labor or property of another.” And it is true that as a matter of California law those are the only two essential elements of theft. See Lopez-Valencia v. Lynch, 798 F.3d 863, 870 (9th Cir. 2015) (citing People v. Ashley, 267 P.2d 271, 279 (Cal. 1954)). Under New York law, by contrast, an indictment must also “assert[] facts supporting every element of the offense charged.” N.Y. Crim. Proc. Law § 200.50(7)(a) (McKinney 2016). The pleading requirement that an indictment must include such supporting facts does not transform them into elements of the charged offense. See People v. Grega, 531 N.E.2d 279, 283–84 (N.Y. 1988) (stating the government is required to allege “a description of the conduct that resulted in the victim’s death” though it is “not an element of the crime”). Perhaps this third textual indicator is more consistently accurate as a rule of exclusion rather than as a rule of inclusion—i.e., that any fact that statutorily need not be included in the indictment cannot be an element of the offense. Cf. 42 Cecily Fuhr et al., Corpus Juris Secundum: Indictments § 168, Westlaw (database updated Sept. 2017) (“An indictment, information, or complaint is generally required to set forth the elements of the offense sought to be charged.”). First, if alternatives carry different punishments, then they must be alternative elements to comply with Apprendi. 134 Mathis, 136 S. Ct. at 2256. In United States v. Lopez–Jacobo, the defendant appealed the district court’s use of the modi­fied categorical approach. 135 656 F. App’x 409, 413 (10th Cir. 2016). The Tenth Circuit concluded that the mod­ified categorical approach was proper, explaining that “[b]ecause Illinois’ statutory alternatives carry different punishments, the subsec­tions reflect alternative elements.” 136 Id. at 414. Similarly, in Singh v. Attorney General, the Third Circuit found a drug statute divisible based on the type of drug, in part because different drugs carried different penalty ranges. 137 839 F.3d 273, 282 (3d Cir. 2016).

Second, if a statute offers only “illustrative examples,” then those examples are mere means rather than elements. 138 Mathis, 136 S. Ct. at 2256. Courts of appeals have also used this tool to determine that statutes list alternative means rather than elements. In United States v. Ritchey, the Sixth Circuit consid­ered prior convictions under a Michigan statute that criminalized breaking and entering into “a tent, hotel, office, store, shop, warehouse, barn, granary, factory or other building, structure, boat, ship, shipping container, or railroad car.” 139 840 F.3d 310, 315 (6th Cir. 2016) (quoting Mich. Comp. Laws Ann. § 750.110 (West Supp. 2016)). Only if the defendant were convicted of burglary of a “building or other structure” could the prior convictions serve as ACCA predicates. 140 Id. (citing Taylor v. United States, 495 U.S. 575, 598 (1990)). Emphasizing the word “other,” the panel reasoned that the statute did not set out alternative elements but rather “non-exhaustive examples” of means that could fulfill a single locational element. 141 Id. at 320 (internal quotation marks omitted) (quoting United States v. Howard, 742 F.3d 1334, 1348 (11th Cir. 2014)). The Seventh Circuit conducted a strikingly similar analysis in United States v. Edwards, concluding that “[t]he statute’s text and struc­ture suggest that the components of each subsection are merely ‘illustra­tive examples’ of particular location types.” 142 836 F.3d 831, 837 (7th Cir. 2016) (quoting Mathis, 136 S. Ct. at 2256). In Edwards, this textual analysis was particularly important because there was no controlling state precedent on the issue of whether the particular location burgled was an element or a means. 143 Id. at 836.

Between state cases and the text of the statute, there are adequate resources to make a nonarbitrary means–element determination in most cases under Mathis. Though the preceding paragraphs have been framed in terms of Fallon’s analytical-bite prong, 144 Fallon, supra note 99, at 1287. the same features of the Mathis tools that allow them to provide nonarbitrary criteria for judicial decisionmaking also produce substantial predictability and consistency. Though the cate­gorical approach is complicated, 145 See United States v. Mayer, 560 F.3d 948, 952 (9th Cir. 2009) (Kozinski, J., dissenting from the denial of reh’g en banc) (describing the categorical approach as “[c]omplex, to be sure”). it is also relatively rule-bound. 146 See United States v. Redrick, 841 F.3d 478, 482 (D.C. Cir. 2016) (describing the categorical approach as “normally rather mechanical”). This has continued to be true in the wake of Mathis. Only about twelve percent of court of appeals cases confronting the divisibility question after Mathis have produced dissents. 147 See infra Appendix. This is a slightly higher dissent rate than the eight pecent rate found in a random sample of published court of appeals deci­sions. 148 See Epstein, Landes & Posner, supra note 107, at 255–56. This should be treated as a rough comparison, as the cases confronting divisibility in Mathis differ from the broader sample in a number of ways having nothing to do with the predictability of applying the Mathis tools. To mention just a few, the divisibility cases are not all selected for publication, are all from 2016 or 2017, are all criminal cases, and are all merits decisions rather than procedural terminations. Such sampling differences matter, as illustrated by the fact that the dissent rate for all court of appeals cases from 1990 to 2007 is just 2.7%. Id. at 265. Nonetheless, the relatively high level of agreement among court of appeals judges in divisibility cases reveals that, in most cases, the Mathis tools produce consistent and predictable results.

Finally, the means–element inquiry satisfies Fallon’s third practical indicator of workability because it does not require empirical judgments beyond courts’ competence. On the contrary, it is an interpretive inquiry that judges are uniquely qualified to undertake. 149 See supra note 131 and accompanying text. Put another way, Mathis requires only that courts resolve a question of law, which raises no concerns about courts’ institutional competence. 150 See Rebecca Sharpless, Toward a True Elements Test: Taylor and the Categorical Analysis of Crimes in Immigration Law, 62 U. Miami L. Rev. 979, 985–93 (2008) (employing the law–fact distinction to defend the practice of allowing adjudicators to determine the legal fact of a defendant’s prior conviction while barring them from finding historical facts about the conduct involved in the conviction); cf. Abram Chayes, The Role of the Judge in Public Law Litigation, 89 Harv. L. Rev. 1281, 1309 (1976) (noting the challenge for judges posed by cases that require “legislative and predictive factfinding”). Thus the Mathis rule satisfies each of Fallon’s practical indicia of workability. 151 This Note does not take up the next step that Fallon’s analysis suggests—that is, whether the all-things-considered benefits of the rule are justified. See Fallon, supra note 99, at 1293–96. Indeed, the thrust of the dissenters’ arguments against the Mathis rule was not that it is totally unworkable but rather that the costs of applying it are not worth the benefits that it produces. See, e.g., Mathis v. United States, 136 S. Ct. 2243, 2259 (Breyer, J., dissenting) (expressing concern that the Mathis rule would “unnecessarily complicate federal sentencing law”). Apprendi and Almendarez-Torres may not allow judicial fact-finding, no matter the practical benefits. In any event, Mathis conclusively resolved the current rule, so this Note focuses instead on how the rule can best be applied.

B. Cases Omitting a Full Discussion of Divisibility

On the whole, the circuit courts have taken Mathis seriously and faithfully applied its instructions. There are a few cases, however, in which the opinion makes a conclusory statement that a statute is divisible based on its text alone. In United States v. Madkins, for example, a Tenth Circuit panel asserted, without analysis, that Kansas’s controlled-substances statute was divisible. 152 866 F.3d 1136, 1145 (10th Cir. 2017) (“Section 609.582, subd. 3, is divisible . . . .”). Though it seems that the defense did not explicitly argue that the statute was indivisible, 153 See Appellant’s Opening Brief and Required Attachments at 17, Madkins, 866 F.3d 1136 (No. 15-3299) (framing its argument as valid “[e]ven if these statutes are divisible”). the absence of more detailed element–means reasoning is surprising given the text of the statute, which listed alternatives in a single block:

[I]t shall be unlawful for any person to sell, offer for sale or have in such person’s possession with intent to sell, deliver, or distribute; prescribe; administer; deliver; distribute; dispense or compound any opiates, opium or narcotic drugs, or any stimu­lant. 154 Madkins, 866 F.3d at 1145 (alteration in original) (quoting Kan. Stat. Ann. § 65-4161(a) (2001)).

Without knowing more about Kansas law, it is not obvious whether this statute lists alternative means or separate crimes with separate ele­ments. Indeed, other cases confronting similar issues have provided fuller discussions of state law. 155 [1].  See, e.g., Chang-Cruz v. Attorney Gen., 659 F. App’x 114, 118 (3d Cir. 2016) (examining state cases and model jury instructions to determine whether “distribution” and “dispensing” are alternative means or alternative elements under New Jersey law); see also Swaby v. Yates, 847 F.3d 62, 67–68 (1st Cir. 2017) (examining state cases, the statutory text, and the indictment to determine whether Rhode Island controlled-substance offenses are divisible on the type of drug). Perhaps the panel’s decision not to ex­amine state law in any detail is understandable because it ultimately concluded that, even under the modified categorical approach, the prior conviction at issue did not qualify as a generic federal “controlled substance offense” because Kansas defined “‘sale’ to include an ‘offer to sell.’” 156 Madkins, 866 F.3d at 1145–48. Whether or not Madkins is correct about the divisibility of Kansas’s statute, 157 The Kansas pattern jury instructions that were current at the time of Madkins’s prior conviction illustrate both the value and limitations of such instructions in the means–element inquiry. For example, the pattern instructions provided two separate entries for offenses under Kan. Stat. Ann. § 65-4161. The first instruction charged an offense involving distribution, sale, and a number of similar terms. Kan. Judicial Council Advisory Comm. on Criminal Jury Instructions, Pattern Instructions for Kansas—Criminal § 67.13-B (3d ed. Supp. 2001),
Archived%20publications/PIK%20Crim/PIK%20Crim%20Supp%202001.pdf [
7F45-7A9S]. The second instruction charged an offense involving possession with intent to sell or offering to sell. Id. § 67.13-C. This bifurcation at least suggests that these two forms of the offense would not be charged together and thus had different elements. Within each instruction, however, there are further alternatives. For example, the second instruction begins, “The defendant is charged with the crime of unlawfully (possessing) (offering to sell) [insert name of narcotic drug or stimulant] with intent to (sell) (deliver) (distribute).” Id. Are these parenthetical terms elements or means? The face of the instructions does not definitively answer that question. Cf. United States v. Titties, 852 F.3d 1257, 1281 (10th Cir. 2017) (Phillips, J., dissenting) (arguing that similar parentheses in Oklahoma’s pattern instructions indicate they contain elements).
its omission of the full analysis that Mathis would re­quire illustrates that the work of applying Mathis can be somewhat bur­densome and ultimately irrelevant to the outcome of the case.

Similarly, in United States v. Lara-Martinez, the Fifth Circuit asserted, without discussion, that “[t]he modified categorical approach is appro­priate because this statute has ‘multiple alternative elements.’” 158 836 F.3d 472, 475–76 (5th Cir. 2016) (citing Mathis v. United States, 136 S. Ct. 2243, 2249 (2016); United States v. Fierro–Reyna, 466 F.3d 324, 327 (5th Cir. 2006)). This lack of analysis is more concerning because the defendant’s brief specifi­cally argued that the statute did not set out alternative elements. 159 Brief for Appellant at 15–16, Lara-Martinez, 836 F.3d 472 (No. 15-41497), 2016 WL 389865; see also Sheridan Green, The Fifth Circuit Holds that Missouri Sexual Abuse of a Minor Is “Crime of Violence” Under the U.S. Sentencing Guidelines, Sheridan Green Law PLLC: Blog (Oct. 12, 2016),
United-States-v-Lara-Martinez-No-15-41497-5th-Cir-Sept-6-2016 [] (“It is interesting that the Court, citing Mathis, resorted to the modified categorical approach without any analysis of whether a Missouri jury would be required to agree unanimously whether the defendant violated [one of the alternative provisions in the statute].”).
Lara-Martinez argued that his prior Missouri conviction for sexual misconduct involving a child did not qualify as a prior conviction for “sexual abuse of a minor,” reasoning that the Missouri statute allowed conviction for con­duct that did not involve an actual minor but rather a law enforcement officer pretending to be a minor. 160 See Lara-Martinez, 836 F.3d at 474–76. The panel rejected this argument, finding first that the statute was divisible, and second that the provision under which Lara-Martinez was charged categorically required abuse of an actual minor. 161 Id. at 475–77. Again, whether or not the court’s divisibility determi­nation is correct, it is striking that it cited Mathis without examining state precedent or statutory text, which the Supreme Court identified as the preferred tools for identifying whether a statute is di­visible. 162 Compare id. at 475–76 (citing Mathis, 136 S. Ct. at 2249), with Mathis, 136 S. Ct. at 2256–57 (explaining how a sentencing court should approach the “threshold” divisibility inquiry). As in Madkins, the missing analysis provides some indication that it can be burdensome to perform. 163   [1]. For a further example of a case that gives short shrift to the means–element distinction, see United States v. Mata, 869 F.3d 640 (8th Cir. 2017). The panel there concluded that a statute was divisible based largely on a state case stating the offense “can be committed by” any of three different acts. Id. at 643 (internal quotation marks omitted) (quoting State v. Leake, 699 N.W.2d 312, 323–24 (Minn. 2005)). That state case, however, does not expressly distinguish between means and elements; in fact, it describes the “essential element” as “force or coercion,” Leake, 699 N.W.2d at 324 (emphasis added), thus suggesting that those two alternatives may be mere means.

These cases highlight the downside of the Mathis approach: Wheth­er a prior conviction qualifies as a predicate offense cannot be deter­mined from the defendant’s record of conviction alone. Rather, because the sentencing judge must determine whether the statute of conviction set out alternative elements or means, getting the right result requires examination of state law and textual analysis of the statute itself. The Descamps majority expressly claimed that looking beyond the record would not be necessary because the “indictment, jury instructions, plea colloquy, and plea agreement—would reflect the crime’s elements. So a court need not parse state law [to determine whether a statute lists ele­ments or means] . . . .” 164 Descamps v. United States, 133 S. Ct. 2276, 2285 n.2 (2013). But Mathis changed course, instead requiring sentencing judges to look to “authoritative sources of state law.” 165 Mathis, 136 S. Ct. at 2256. This necessarily increases the research burden on parties and the courts at­tempting to apply the categorical approach. 166 See id. at 2264 (Breyer, J., dissenting) (“That research [into whether a statute sets out alternative means or elements] will take time and is likely not to come up with an answer.”); Fogg Transcript of Sentencing Hearing, supra note 132, at 13 (“It’s a really hard issue to research because it gets tied up in the individual state statutes, so it’s hard to kind of use Westlaw and Lexis to research this.”). Some federal judges have, understandably, continued to object to this consequence of Mathis. 167 See, e.g., United States v. Martinez-Lopez, 864 F.3d 1034, 1058 (9th Cir. 2017) (Bybee, J., concurring in part and dissenting in part) (declaring himself “frustrated with the whole endeavor” of the categorical and modified categorical approach); United States v. Tavares, 843 F.3d 1, 19 (1st Cir. 2016) (characterizing the federal law of prior convictions as “a Rube Goldberg jurisprudence of abstractions piled on top of one another in a manner that renders doubtful anyone’s confidence in predicting what will pop out at the end”); United States v. Doctor, 842 F.3d 306, 313 (4th Cir. 2016) (Wilkinson, J., concurring) (“[T]he purported administrative benefits of the categorical approach have not always worked as advertised. Judges have simply swapped factual inquiries for an endless [gantlet] of abstract legal questions.”); United States v. Edwards, 836 F.3d 831, 838 (7th Cir. 2016) (noting the “practical difficulty that can arise in applying the Mathis/Descamps rule”); Almanza–Arenas v. Lynch, 815 F.3d 469, 482 (9th Cir. 2016) (Owens, J., concurring) (“We should no longer tinker with the machinery of Descamps.”); United States v. Brown, No. 7:02-CR-024, 2016 WL 7441717, at *13 n.14 (W.D. Va. Dec. 23, 2016) (“Given the difficulties inherent in deciphering the distinction between elements and means, the court views the analysis in this case to be anything but, in Mathis’ terms, ‘easy.’” (quoting Mathis, 136 S. Ct. at 2256)); see also Doug Keller, Causing Mischief for Taylor’s Categorical Approach: Applying “Legal Imagination” to Duenas-Alvarez, 18 Geo. Mason L. Rev. 625, 625 (2011) (“The categorical approach . . . has become the rule of perpetuities of criminal law.”); Lee, Future of the Categorical Approach, supra note 23, at 266–68 (“It is undoubtedly true that many, perhaps most, federal judges are confused about this area . . . . [A]nd that confusion may well continue to a significant degree after Mathis.”). These difficulties have likely been pushed to the forefront of judicial attention by the Supreme Court’s decision to make its holding in Johnson retroactive on collateral review. See Welch v. United States, 136 S. Ct. 1257, 1265 (2016). That decision alone increased the number of cases against the United States by fifty-five percent during the federal judiciary’s last reporting year. Chief Justice John Roberts, 2016 Year-End Report on the Federal Judiciary 12 (2016),
2016year-endreport.pdf []. In this large volume of cases, lower courts were sorting through the Mathis rule simultaneously with interrelated questions of retroactivity and habeas corpus review. See, e.g., Holt v. United States, 843 F.3d 720, 721–22 (7th Cir. 2016) (considering the interaction of retroactivity and divisibility issues); Traxler v. United States, No. 1:16-CV-747, 2016 WL 4536329, at *4 (W.D. Mich. Aug. 31, 2016) (noting that these issues arising from Johnson have divided lower courts), vacated, No. 16–2280, 2017 WL 4124880 (6th Cir. Mar. 7, 2017).

It is too early to say whether these critiques will lead to a judicial or legislative rejection of the Mathis rule. 168 See Lee, Future of the Categorical Approach, supra note 23, at 266 (predicting “this disagreement [about workability] will persist for some time”); see also Mathis, 136 S. Ct. at 2258 (Kennedy, J., concurring) (expressing his view that “the elements based approach . . . is required only by the Court’s statutory precedents, which Congress remains free to overturn”). But all should agree that greater clarity and consistency in this area would be valuable. The following sec­tion, therefore, turns to the third interpretive tool that the Mathis major­ity suggested, the record of the prior conviction, 169 See Mathis, 136 S. Ct. at 2256–57. identifying two ways in which its proper application is somewhat unclear.

C. Unresolved Questions in Applying Mathis

Mathis suggested a third tool to employ when “state law fails to pro­vide clear answers.” 170 Id. at 2256. Specifically, judges can take a “peek at the [record] documents” for the “sole and limited purpose of determining whether [the listed items are] element[s] of the offense.” 171 Id. at 2256–57 (alteration in original) (quoting Rendon v. Holder, 782 F.3d 466, 473–74 (9th Cir. 2015) (Kozinski, J., dissenting from the denial of reh’g en banc)). The Court elaborated with two examples of what such a peek might reveal. 172 Id. at 2257. First, if “one count of an indictment and correlative jury instructions” list all the alternatives, then “[t]hat is as clear an indication as any that each alter­native is only a possible means of commission.” 173 Id. Second, the indictment and jury instructions “could indicate, by referencing one alternative term to the exclusion of all others, that the statute contains a list of elements, each one of which goes toward a separate crime.” 174 Id. Then the Court added a crucial caveat, that “such record materials will not in every case speak plainly, and if they do not, a sentencing judge will not be able to satisfy ‘Taylor’s demand for certainty’ when determining whether a de­fendant was convicted of a generic offense.” 175 Id. (quoting Shepard v. United States, 544 U.S. 13, 21 (2005)).

This passage raises two key questions. First, is “certainty” the stand­ard for determining whether a statute is divisible? Second, under what circumstances could the charging of a single statutory alternative provide sufficient basis to conclude that a statute is divisible? Lower-court decisions reveal differing views on these questions, which complicates the application of Mathis.

1. The Role of “Certainty.” — What should we make of “Taylor’s de­mand for certainty”? The language originated in Shepard, in which the Court characterized Taylor as requiring that “evidence of generic convic­tion be confined to records of the convicting court approaching the cer­tainty of the record of conviction in a generic crime State.” 176 Shepard, 544 U.S. at 23. In Mathis, the Court made clear that the modified categorical approach analysis requires two steps: First, determine whether the statute lists alternative elements or alternative means; 177 Mathis, 136 S. Ct. at 2256. then, only if the alternatives are ele­ments, determine whether the defendant necessarily admitted elements sufficient to categorically qualify as a predicate offense. 178 Id. at 2254–56. In Shepard, the “demand for certainty” language was applied at this second stage of modified categorical approach analysis, 179 See Shepard, 544 U.S. at 26 (making no mention of the means–elements distinction in its holding). but the language in Mathis can be read to apply to either stage, or both. On one hand, the Mathis opin­ion adverts to the “demand for certainty” in a section devoted to answer­ing the means–element question. 180 The section begins, “The first task for a sentencing court faced with an alternatively phrased statute is thus to determine whether its listed items are elements or means.” Mathis, 136 S. Ct. at 2256. On the other hand, the sentence in which Mathis quoted the language does not refer directly to the means–element distinction; it refers to the broader task of “determining whether a defendant was convicted of a generic offense.” 181 Id. at 2257. The following sentence, however, says that “between [the record] documents and state law . . . indeterminacy should prove more the exception than the rule.” 182 Id. This language in particular suggests that the import of “cer­tainty” is not limited to use of the record documents but also to the in­quiry into “state law.” Thus there are two possible readings of this language from the Mathis opinion: (1) A court may determine that a statute is divisible only if it is “certain” that it sets out alternative ele­ments rather than means, or (2) certainty is required only in the second stage, when the court is determining what the defendant necessarily admitted.

The Eighth Circuit, in United States v. Horse Looking, took the second approach. 183 See 828 F.3d 744, 747–49 (8th Cir. 2016). Horse Looking argued that his conviction under the statute was not for “a misdemeanor crime of domestic violence” because one of the alternatives listed in the statute did not require the use of force. 184 See id. at 746. The panel concluded that a South Dakota domestic assault statute was divisible, without invoking the “demand for certainty.” 185 See id. at 747–48. During his plea colloquy, Horse Looking admitted that he had pushed his wife, and the victim had testified she had abrasions as a result. 186 Id. The panel concluded that this record was consistent with a conviction under either a subsec­tion of the statute requiring the use of violent force or a subsection re­quiring only “attempting by physical menace to put another in fear of imminent bodily harm.” 187 Id. at 748. The panel then quoted Mathis’s language about the demand for certainty and concluded that this record could not satisfy that demand:

It is clear that Horse Looking admitted using physical force against his wife, and that he could have been found guilty of a crime that has, as an element, the use of force against his wife. But the judicial record does not establish that Horse Looking necessarily was convicted of an assault that has the required ele­ment. He was charged in the alternative with a non-qualifying assault, and the state court did not specify which alternative was the basis for conviction. 188 Id. at 749.

This case certainly illustrates that the categorical approach can be blind to the actual conduct that produced a conviction, but that is by design. 189 See Descamps v. United States, 133 S. Ct. 2276, 2293 (2013) (“The modified approach does not authorize a sentencing court to substitute such a facts-based inquiry for an elements-based one.”). For present purposes, the important takeaway is that the panel in Horse Looking did not apply the “demand for certainty” to its divisibil­ity inquiry but only to its determination of whether the record showed the defendant necessarily admitted a certain statutory alternative.

The Sixth Circuit, however, has invoked the “demand for certainty” in conducting a divisibility inquiry. In United States v. Ritchey, the defend­ant argued that his Michigan breaking-and-entering convictions did not qualify as ACCA predicates because the statute criminalized burglary of not just buildings but also, “among other things, tents, boats, and rail­road cars.” 190 840 F.3d 310, 315 n.1 (6th Cir. 2016). After concluding that state cases and the statute’s text indi­cated that the alternative locations were means rather than elements, the panel found that consideration of the record documents produced the same conclusion. 191 Id. at 319–21. Though Ritchey’s charging documents did allege the specific locations that he broke into, the panel noted that other portions of the record suggested that the alternatives were mere means—for example, one charged entry of a “BARN/GARAGE,” thus suggesting that the jury need not find that Ritchey broke into either specific location beyond a reasonable doubt. 192 Id. at 321. Characterizing the record documents as “at the very most, inconclusive,” the panel concluded that they could not satisfy Taylor’s demand for certainty. 193 Id. The Ritchey opinion exemplifies an approach that treats the “demand for certainty” as relevant to the di­visibility inquiry but only insofar as the record documents are used as evidence of divisibility. When parsing the statute and state-court interpretations thereof, the Sixth Circuit treated divisibility as an ordinary question of statutory interpretation not dependent on a finding of “certainty.” 194 See id. at 318–20.

The clearest illustration of differing applications of the “demand for certainty” comes from an Eleventh Circuit case, United States v. Gundy. 195 842 F.3d 1156 (11th Cir. 2016). There, the panel majority concluded that the text of Georgia’s burglary statute and state-court interpretations indicated that the statute was di­visible. 196 Id. at 1166–68. The majority further reasoned that, even if state law were am­biguous, the record of Gundy’s indictments was sufficient to “satisfy Taylor’s demand for certainty” because they alleged that he had burgled a “dwelling house” and a “business house.” 197 Id. at 1170. In dissent, Judge Jill Pryor first reached the opposite conclusion about Georgia law. 198 Id. at 1172–77 (Pryor, J., dissenting). Then her opinion turned to the record materials and observed that some of the indictments charged entry into a “business house,” a term not found in the text of the statute. 199 Id. at 1178. Judge Pryor reasoned that a term not present in the text of the statute cannot be an element and said the majority had “misconceive[ed] the appropriate inquiry under Mathis at this stage of the analysis” by failing to frame the test in terms of whether the record showed the alternatives to be elements or means. 200 Id. at 1179. Importantly for the present discussion, Judge Pryor then characterized the “‘demand for cer­tainty’” language in Mathis as requiring that statutes be found indivisible whenever “state law and the records of a conviction are inconclusive re­garding a statute’s divisibility.” 201 Id. (quoting Mathis v. United States, 136 S. Ct. 2243, 2257 (2016)). Professor Evan Lee also advocates the use of a certainty standard for determining whether a defendant has a qualifying prior conviction 202 Lee, Future of the Categorical Approach, supra note 23, at 276–77 (arguing the modified categorical approach should be abolished if defendants continue, after Mathis, to be “illegally punished for the same crime twice because courts are confused about whether statutes are divisible or indivisible”). and suggests as a descriptive matter that such a certainty standard is what Mathis requires. 203 Evan Lee, Opinion Analysis: Victory for the “Categorical Approach” in Immigration and Federal Criminal Sentencing—But for How Long?, SCOTUSblog (June 24, 2016),
approach-in-immigration-and-federal-criminal-sentencing-but-for-how-long [
D7BQ-FT9K] (summarizing the Mathis rule as, “If in doubt, it’s out”).
Despite the fact that Mathis’s text invites this reading, the weight of authority does not de­mand “certainty” in making the divisibility inquiry 204 Very few of the court of appeals cases even quote the “demand for certainty” language. And even in Fogg, a case in which the judges disagreed about the proper reading of state law, the antidivisibility dissent did not invoke any “certainty” requirement, though it would have certainly strengthened the argument against finding divisibility. See United States v. Fogg, 836 F.3d 951, 960–62 (8th Cir. 2016) (Bright, J., dissenting). but only in the use of the record documents.

2. Record Documents as Proof of Divisibility. — As the above cases illustrate, lower courts have read the “certainty” language in Mathis in a few different ways, each of which might produce different results. Part III takes up which of these readings is most consistent with the Court’s ju­risprudence. But Gundy also highlights the second key question that fol­lows from Mathis: When can the record documents themselves provide sufficient basis to conclude that a statute is divisible? Mathis provided three hypothetical situations in which the record documents would help answer the means–element question, two of which indicate the statute is indivisible and one of which suggests that it is divisible. The table below summarizes this guidance, along with the case in which the record is unhelpful. 205 Mathis, 136 S. Ct. at 2257 (quoting Shepard v. United States, 544 U.S. 13, 21 (2005)).

Table 1: Mathis’s Guidance for Using Record Documents

Contents of Record Result
The indictment and jury instructions list each of the alternatives from the statute.
The indictment and jury instructions use a blanket term, like “premises,” that encompasses alternatives.
“That is as clear an indication as any that each alternative is only a possible means of commission, not an element that the prosecutor must prove to a jury beyond a reasonable doubt.”
The indictment and jury instructions reference “one alternative term to the exclusion of all others.” That “could indicate . . . the statute contains a list of elements, each one of which goes toward a separate crime.”
The record materials do not “speak plainly.” “A sentencing judge will not be able to satisfy ‘Taylor’s demand for certainty’ when determining whether a defendant was convicted of a generic offense.”


Two immediate observations are worth making about this text. 206 Judge Pryor makes these two points in Gundy, 842 F.3d at 1177 n.9 (Pryor, J., dissenting). First, the Court states that the record sometimes provides a “clear . . . indication” that a statute is indivisible but that the record at most “could indicate” that a statute is divisible. 207 Mathis, 136 S. Ct. at 2257. This suggests that the record is more powerful as a tool for ruling out divisibility than for finding divisi­bility. Second, each hypothetical assumes that there is both an indict­ment and “correlative jury instructions.” 208 Id. This casts doubt on the persuasiveness of arguments that are based on indictments alone.

In Gundy, the majority seemed to think that the record put the case in the second box above, because each indictment alleged a specific al­ternative in isolation. 209 Gundy, 842 F.3d at 1170 (majority opinion) (“We . . . conclude that the terms ‘dwelling house’ and ‘business house’ [in the indictments] satisfy Taylor’s demand for certainty that Gundy’s convictions were for burglary of a building or other structure, which is a generic burglary.”). By contrast, Judge Pryor concluded that the rec­ord did not speak plainly enough to satisfy the demand for certainty. 210 Id. at 1179 (Pryor, J., dissenting) (“The majority should acknowledge that the two terms found in Mr. Gundy’s indictments—one of which cannot be found in the text of the statute and therefore cannot be an element—provide insufficient clarity to conclude that Georgia’s burglary statute is divisible.”). As discussed above, 211 See supra section II.A.2. most court of appeals cases applying Mathis conclude that the statute’s text and state precedents answer the means–element question, so the stakes attached to reading the record of convic­tion are lower than in Gundy. Nonetheless, many cases use the third Mathis tool as confirmation of the conclusion from analysis of the first two. In Chang-Cruz v. Attorney General, for example, the Third Circuit noted that the judgments of conviction listed both of the alternatives, thus providing “clear . . . indication” that they were means rather than elements. 212 659 F. App’x 114, 118 (3d Cir. 2016) (internal quotation marks omitted) (quoting Mathis, 136 S. Ct. at 2257). Similarly, the Eighth Circuit found that a statute was indi­visible based in part on the fact that the defendant’s charging document used a “single umbrella term.” 213 United States v. McFee, 842 F.3d 572, 575–76 (8th Cir. 2016) (internal quotation marks omitted) (quoting Mathis, 136 S. Ct. at 2257).

Courts have also invoked the record in support of a finding of divis­ibility. The Fifth Circuit, for example, reasoned that a judicial confession reciting the text of only one statutory subsection supported a finding that the defendant’s conviction necessarily involved violation of that subsec­tion. 214 United States v. Uribe, 838 F.3d 667, 669 (5th Cir. 2016); see also Ibanez-Beltran v. Lynch, 858 F.3d 294, 298 (5th Cir. 2017) (concluding that a “plea agreement, judgment, and [model] instructions are enough, without settled state law to the contrary, to hold” that a statute is divisible, when the documents list only a single alternative). And the Tenth Circuit has used the fact that an indictment cited only a particular subsection to bolster its conclusion that the statute’s subsections set out divisible elements. 215 United States v. Lopez–Jacobo, 656 F. App’x 409, 414 (10th Cir. 2016) (“Moreover, ‘an indictment . . . could indicate, by referencing one alternative term to the exclusion of all others, that the statute contains a list of elements.’” (alteration in original) (quoting Mathis, 136 S. Ct. at 2257)). Note the interesting placement of the ellipsis here, original to the Tenth Circuit opinion, omitting the words “and jury instructions.” As Judge Pryor observes, the inclusion of those words in Mathis provides a basis to be skeptical of the claim that an indictment alone can furnish sufficient basis to find divisibility. Gundy, 842 F.3d at 1177 n.9. Finally, the Seventh Circuit opinion in United States v. Edwards provides an example in which the rec­ord was unhelpful in resolving divisibility; the court concluded that “in Wisconsin neither the charging documents nor a plea colloquy will necessarily reflect only the elements of a crime.” 216 836 F.3d 831, 838 (7th Cir. 2016).

To sum up, though Mathis provided some guidance on how to use record documents as part of divisibility analysis, lower courts have not adopted a uniform reading of that guidance. In particular, there is disa­greement about what role “certainty” plays in the divisibility analysis and about when record documents are sufficient to themselves indicate that a statute is divisible. As Gundy illustrates, the answers to these questions are critical for cases in which authoritative sources of state law prove incon­clusive. Part III seeks to articulate solutions that are consistent with Mathis and the Court’s other categorical-approach cases.

III. Clarifying Mathis: How to Use the Record of Prior Conviction

The above exploration produces two main findings about how Mathis has functioned in application. First, its state-law approach to divis­ibility is workable, but takes work. That is, the precedential and textual analysis that Mathis demands has not proven as indeterminate as its de­tractors predicted, but it does require sentencing judges to delve into state-law questions that do not always have obvious answers. Some cases are indeed “easy” because of state decisions that are clearly on point or because the statutory alternatives carry different punishments. But there remain challenging cases in which such decisions and dispositive text are absent. These are the cases in which the “peek” at the record documents Mathis authorizes takes center stage.

The opinion in Mathis, however, provided relatively little guidance on how to conduct this inquiry. For one, it did not precisely explain the relationship between “Taylor’s demand for certainty,” the record docu­ments, and the divisibility inquiry. 217 See Mathis, 136 S. Ct. at 2256–57. Also, by saying only that certain rec­ord documents “could indicate” that the statute sets out alternative ele­ments, 218 Id. at 2257. Mathis did not fully explain when such documents would indi­cate divisibility. This Part advances three interpretive clarifications of Mathis.

A. Treating Divisibility as a Threshold Inquiry—an Unnecessary Burden

Whether a court “peeks” at the record or scrutinizes it fully, it should always see the same thing. Mathis describes the inquiry necessary for the modified categorical approach as a two-step process: (1) Deter­mine whether statutory alternatives are elements or means; (2) only if they are elements, determine which element the defendant was neces­sarily convicted of. 219 Id. at 2256. The Court expressly described the means–element question as a “threshold inquiry” and resolving it as “[t]he first task for a sentencing court.” 220 Id. This section argues that, once one has exhausted the first two Mathis tools, this bifurcation is actually misleading about the analysis required and suggests an order-of-decision rule that imposes un­necessary analytical burdens on sentencing judges.

To see why the bifurcation is confusing, it is helpful to define pre­cisely what the second step of the inquiry is. Mathis says that, once the sentencing court has determined that the alternatives are elements, the second step is to “review the record materials to discover which of the enumerated alternatives played a part in the defendant’s prior convic­tion, and then compare that element (along with all others) to those of the generic crime.” 221 Id. When employing the first two Mathis tools, this distinction is comprehensible. For example, one might find state prece­dent that a statute is divisible but that the record documents in the case do not identify which of the alternatives the defendant was actually con­victed of. That, for example, is what the Third Circuit concluded in Singh v. Attorney General. At issue there was whether Singh’s prior conviction under Pennsylvania’s controlled-substances law qualified as an aggravat­ed felony for immigration purposes. 222 Singh v. Attorney Gen., 839 F.3d 273, 277–78 (3d. Cir. 2016). The panel found state precedent demonstrating that the Pennsylvania statute was divisible on the type of drug involved. 223 Id. at 283–84. However, because the charging documents in Singh’s record did not actually identify the type of drug, the court concluded that, even applying the modified categorical approach, his prior convic­tions were not federal-law aggravated felonies. 224 Id. at 284–86. In such a case, one can logically find that the statute is divisible, but that the prior conviction was not for a qualifying offense. The same is true when the court is perform­ing textual analysis of divisibility. 225 See United States v. Tavares, 843 F.3d 1, 18–20 (1st Cir. 2016) (finding the statute divisible, but remanding for consideration of other issues, including whether the record documents show that the defendant was convicted under the qualifying alternative).

The language of Mathis makes clear that the Court intends this dis­tinction to persist when examining the record documents. Indeed, the Court says that the “‘peek at the [record] documents’ is for ‘the sole and limited purpose of determining whether [the listed items are] element[s] of the offense.’” 226 Mathis, 136 S. Ct. at 2256–57 (2016) (quoting Rendon v. Holder, 782 F.3d 466, 473–74 (9th Cir. 2015) (Kozinski, J., dissenting from the denial of reh’g en banc)). This “peek” remains a threshold inquiry, as indicated by the Court’s language that “only if [the record documents show the alternatives are elements] can the court make further use of the materials.” 227 Id. at 2257.

But when one considers the Court’s three hypotheticals, it becomes apparent that this bifurcation makes no practical difference. In the first hypothetical, the indictment and jury instructions list all the alternatives listed in the statute. Though it is true that this is “clear indication” that the statute is indivisible, it is equally clear evidence that divisibility is ir­relevant, because the record furnishes no basis for determining which of the alternatives the defendant was actually convicted of. The same is true in the second hypothetical, involving record documents that use only an “umbrella term.” The Court is surely right that such documents show the statute is indivisible, but—because they do not identify a particular alternative—they could not justify applying a prior-conviction enhancement even if the statute were divisible. In these hypotheticals, therefore, the “peek” can only ever produce the same result as a longer look.

The same is true of the third hypothetical peek at the record docu­ments, in which they “referenc[e] one alternative term to the exclusion of all others.” 228 Id. If a sentencing judge concludes that such exclusive spec­ification is indeed sufficient to indicate that the listed alternatives are elements rather than means, she will have necessarily found the answer to Mathis’s second step because the record indicates that the prior con­viction was based only on the specified alternative.

If peeking at the record documents and using them to conduct the full modified categorical approach can only ever produce the same re­sult, then why distinguish the two steps? The original suggestion for a “peek” at the record documents came from Judge Kozinski’s opinion dissenting from the denial of rehearing en banc in Rendon v. Holder. 229 See id. at 2256–57 (quoting Rendon, 782 F.3d at 473–74). There, Judge Kozinski was trying to make sense of a footnote in Descamps that had read, “Whatever a statute lists (whether elements or means), the [Shepard] documents . . . reflect the crime’s elements.” 230 Rendon, 782 F.3d at 473 (quoting Descamps v. United States, 133 S. Ct. 2276, 2285 n.2 (2013)). Though seem­ingly endorsing this portion of Judge Kozinski’s opinion, Mathis is in some respects flatly inconsistent with the approach Judge Kozinski out­lined. First, Judge Kozinski thought that the “peek” at the record docu­ments provided a path to avoid “the laborious and often inscrutable exercise of parsing state law.” 231 Id. at 474.  Such parsing, of course, is precisely what Mathis requires. 232 See Mathis, 136 S. Ct. at 2256 (requiring inquiry into “authoritative sources of state law”).

Second, Judge Kozinski described the second step as an examination of the record documents to determine “whether a defendant committed a state crime falling within the ambit of the relevant federal statute.” 233 Rendon, 782 F.3d at 473. Mathis, however, shows that even the modified categorical approach does not permit inquiry into the real-world fact of what crime the defendant “committed.” Rather, as the Court explained,

[T]he modified approach serves—and serves solely—as a tool to identify the elements of the crime of conviction when a statute’s disjunctive phrasing renders one (or more) of them opaque. It is not to be repurposed as a technique for discovering whether a defendant’s prior conviction, even though for a too-broad crime, rested on facts (or otherwise said, involved means) that also could have satisfied the elements of a generic offense. 234 Mathis, 136 S. Ct. at 2253–54 (citation omitted).

The modified categorical approach thus turns on conviction, not commission. With that understanding, a “peek” at the record documents can never produce a different result than the most detailed scrutiny of those documents. This follows from the fact that the second-stage inquiry is limited to elemental facts; that limitation renders the first stage of the inquiry superfluous.

If the “peek” at the record does not ever change the ultimate result, why require sentencing judges to take that step first? A better way of un­derstanding Mathis’s guidance about the use of the record is as a reaffir­mation of the limit on judicial fact-finding. One potential benefit is that it forces sentencing judges to expressly consider which facts of the prior conviction are elemental, thus serving to discourage the temptation of judicial fact-finding. But characterizing the means–element distinction as a threshold inquiry also has a significant downside. Specifically, it suggests that courts cannot dispose of cases by simply saying that, no matter whether a statute’s alternatives are elements or means, the defendant’s record does not show which alternative was involved in the prior conviction.

Treating the means–element distinction as a threshold question is somewhat akin to an approach that the Court rejected in the qualified immunity context. In Saucier v. Katz, the Court had required that courts deciding upon qualified immunity defenses must first determine whether the plaintiff has alleged facts that would establish violation of a constitu­tional right and only then determine whether that constitutional right was clearly established. 235 533 U.S. 194, 201 (2001). Eight years later, the Court receded from that rule, instead granting lower-court judges the discretion to decide the case on the second prong alone. 236 Pearson v. Callahan, 555 U.S. 223, 236 (2009). The Court reasoned that requiring the threshold determination “sometimes results in a substantial expendi­ture of scarce judicial resources on difficult questions that have no effect on the outcome of the case.” 237 Id. at 236–37.

Much the same is true of the means–element distinction in Mathis. As explored above, the means–element issue is not always easy to re­solve. 238 See supra notes 124–130 and accompanying text. But the requirement in Mathis that courts treat it as a threshold inquiry may contribute to decisions like Horse Looking, in which the court provided cursory analysis of divisibility and concluded that the record of conviction ultimately provided an insufficient basis to reveal which of the statutory alternatives the defendant was convicted of, even having found that they were elements. 239 See United States v. Horse Looking, 828 F.3d 744, 748 (8th Cir. 2016). If not for the instruction in Mathis to treat the means–element distinction as a threshold inquiry, Horse Looking could have simply said nothing about divisibility and resolved the case by ob­serving that the record documents were inconclusive.

A better reading of Mathis, therefore, is as holding (1) that a sen­tencing court cannot conclude that a prior conviction involved a certain fact without first finding that fact to be elemental, but (2) that a sentenc­ing court can conclude that a prior conviction did not include that fact by finding either that the fact is legally non-elemental or not necessarily found in the defendant’s record. Indeed, a number of opinions have ex­pressly declined to decide divisibility when the record documents do not speak clearly enough to support an enhancement. 240 United States v. Montanez-Trejo, No. 16-41088, 2017 WL 3887991, at *4 (5th Cir. Sept. 5, 2017) (“We need not decide whether the Nebraska statute at issue here is divisible because we conclude that, even if the district court did not plainly err in finding that it is divisible, the modified categorical approach does not clarify the subsection under which Montanez–Trejo was convicted.”); United States v. Guillen-Cruz, 853 F.3d 768, 771 (5th Cir. 2017) (“[B]ecause Guillen-Cruz’s prior offense is not an aggravated felony under either [the categorical or modified categorical] approach, we pretermit deciding which approach is applicable.”); United States v. Arriaga-Pinon, 852 F.3d 1195, 1199 (9th Cir. 2017) (“[W]e need not reach [the question of divisibility] in this case because, even assuming [the statute is divisible], the conviction fails to satisfy the modified categorical test at stage three, and therefore is not a qualifying predicate offense.”); United States v. Driver, 663 F. App’x 915, 919 (11th Cir. 2017) (“We need not [decide divisibility] because . . . there are no Shepard documents in the record that would enable us to apply the [modified categorical] approach to the particular false imprisonment conviction at issue in this case.”).

B. The Proper Scope of “Certainty”

A second question explored above is the meaning of the “certainty” language in Mathis. Court of appeals cases have offered three interpreta­tions: (1) require that the divisibility of a statute be certain before mov­ing to the second step of Mathis; 241 See United States v. Martinez-Lopez, 864 F. 3d 1034, 1056 (9th Cir. 2017) (Berzon, J., dissenting); United States v. Sykes, 864 F.3d 842, 844 (8th Cir. 2017) (Colloton, J., dissenting from denial of reh’g en banc) (“[A]n inconclusive inquiry means that the prior convictions do not qualify, and the sentencing enhancement does not apply.”); United States v. Titties, 852 F.3d 1257, 1272 n.19 (10th Cir. 2017) (“[U]ncertainty [about state law] favors [the defendant] because the Government bears the burden of proving a prior conviction qualifies under the ACCA, and we do not count a prior conviction if its ACCA qualification is suspect.” (citation omitted) (citing United States v. Delossantos, 680 F.3d 1217, 1219 (10th Cir. 2012))); Mathis v. United States, 136 S. Ct. 2243, 2257 (2016))); United States v. Gundy, 842  F.3d 1156, 1179 (11th Cir. 2016) (Pryor, J., dissenting). (2) require certainty in the exercise of using record documents to inform divisibility; 242 See United States v. Ochoa, 861 F.3d 1010, 1018 (9th Cir. 2017) (invoking “certainty” only when examining record documents); United States v. Ritchey, 840 F.3d 310, 321 (6th Cir. 2016) (same). and (3) require certainty only in the second step, when determining whether a prior conviction necessarily rested on a particular element of a divisible statute. 243 See Martinez-Lopez, 864 F.3d at 1043 (en banc); United States v. Horse Looking, 828 F.3d 744, 748–49 (8th Cir. 2016). As the preceding section shows, approaches (2) and (3) are functionally identi­cal—if the record documents of a prior conviction are sufficiently certain to show that the statute was divisible, then they are necessarily certain enough to show that defendant’s conviction rested on a particular statu­tory alternative.

Therefore, the only meaningful choice is between that approach and requiring certainty in the divisibility inquiry as a whole. The place­ment of the language in Mathis could support either approach. An en banc Ninth Circuit case, United States v. Martinez-Lopez, 244 864 F.3d 1034. illustrates how these two possibilities operate in practice. The controlled-substances statute there at issue prohibited a variety of acts including “importation, sale, furnishing, administration, etc.” 245 Id. at 1041 (quoting People v. Patterson, 778 P.2d 549, 556 (Cal. 1989)). The majority concluded that California law indicated the statute was divisible on the actus reus requirement, without mentioning the demand for certainty. 246 Id. at 1041–43. Then, in applying the modified categorical approach, the majority noted that during the plea colloquy, Martinez-Lopez had been asked, “[O]n or about December 31st, 1997, [did] you . . . sell cocaine base—.42 grams of cocaine base?” 247 Id. at 1043 (alteration in original). He responded, “Yes.” 248 Id. The majority concluded, “Based on this exchange, we can say—with the certainty that Taylor demands”— that Martinez-Lopez had been convicted for selling cocaine. 249 Id. (citing Mathis v. United States, 136 S. Ct. 2243, 2257 (2016)).

Judge Marsha Berzon, dissenting from this portion of the majority’s opinion, adopted a different approach. In her view, “[D]etermining whether a disjunctively worded statute refers to alternative elements or alternative means is subject to the Court’s more general ‘demand for certainty when identifying a generic offense.’” 250 Id. at 1046 (Berzon, J., concurring in part and dissenting in part) (quoting Shepard v. United States, 544 U.S. 13, 21–22 (2005)) (citing Mathis, 136 S. Ct. at 2257). Further, she argued that the majority opinion “ignore[ed] the Court’s repeated direction to focus only on what must be admitted or proven beyond a reasonable doubt to sustain a conviction.” 251 Id.; see also id. at 1056 (“As I understand the line of cases culminating in Mathis, the certainty requirement cuts in a specific direction: Where there is indeterminacy after all the modes of inquiry prescribed in Mathis are exhausted, a federal court must treat the state statute as indivisible . . . .” (citing Mathis, 136 S. Ct. at 2257)). Judge Berzon’s opinion read Mathis to require that state law or the record “provide a definitive answer” to the divisibility inquiry. 252 Id. at 1048, see also id. at 1046 (“Our inquiry is over if ‘a state court decision definitively answers the question’ . . . .” (emphasis added) (quoting Mathis 136 S. Ct. at 2256)). In short, the majority demanded factual certainty about whether the defendant admitted qualifying conduct, but not legal certainty about whether the statute was divisible. Judge Berzon’s opinion would have demanded legal certainty as a prerequisite to reaching the modified categorical approach.

The majority has the better of this methodological debate for three reasons. First, as a precedential matter, the “demand for certainty” language that Mathis cites originally appeared in a context discussing the propriety of using record documents to determine the elements of a prior conviction, not a broader inquiry into the divisibility of state law. 253 See Mathis, 136 S. Ct. at 2257 (quoting Shepard, 544 U.S. at 21). In Shepard, the sentence in which the “certainty” language originally appeared read: “[T]he Government pulls a little closer to Taylor’s demand for certainty when identifying a generic offense by emphasizing that the records of the prior convictions used in this case are . . . free from any inconsistent, competing evidence on the pivotal issue of fact separating generic from nongeneric burglary.” Shepard, 544 U.S. at 21–22. Further, Mathis said that “between [the record] documents and state law . . . indeterminacy should prove more the exception than the rule.” 254 Mathis, 136 S. Ct. at 2257. If the standard for divisibility is “certainty,” this prediction from Mathis would turn out to be incorrect; courts would be required to find far more statutes indivisible due to uncertain state law. Though many courts of appeals have found guidance in state cases, those cases do not often “definitively” resolve the means–element inquiry. 255 See supra note 105 (noting the wide variety of legal contexts in which state cases have arisen). Second, as a theoretical matter, whether a statute is divisible is a legal, not factual, determination. 256 In some cases, distinguishing law from fact can be “vexing,”  Pullman-Standard v. Swint, 456 U.S. 273, 288 (1982) (citing Baumgartner v. United States, 322 U.S. 665, 671 (1944)). But divisibility is a pure question of statutory interpretation; resolving such a question produces what is archetypally “‘law’—conclusions about the existence and content of governing legal rules, standards, and principles.” Henry P. Monaghan, Constitutional Fact Review, 85 Colum. L. Rev. 229, 235 (1985); see also Douglas A. Berman, Conceptualizing Blakely, 17 Fed. Sent. R. 89, 92 (noting the presence of issues of pure law and pure fact in criminal sentencing). Thus it does not raise the specter of judicial fact-finding in the way that does a resort to the record documents. 257 The fact–law distinction often turns on whether “one judicial actor is better positioned than another to decide the issue in question.” Miller v. Fenton, 474 U.S. 104, 114 (1985); see also United States v. Andrews, 808 F.3d 964, 969 (4th Cir. 2015) (“Issues of law do often arise in sentencing, and the standard of review for such issues is obviously de novo.”). The prevalence of guilty pleas means the best-positioned actor, the state-court judge in the prior case, will probably not have resolved the divisibility issue. But the absence of clearly settled state law does not render federal courts incompetent to apply it when necessary. See United States v. Reyes, 866 F.3d 316, 321 n.4 (5th Cir. 2017) (“When, as here, a federal court must identify and apply state law in the absence of a clearly controlling state supreme court opinion, the analogous Erie inquiry calls on federal courts to ‘guess’ how a state supreme court ‘would decide.’” (quoting Howe ex rel. Howe v. Scottsdale Ins. Co., 204 F.3d 624, 627 (5th Cir. 2000))). Courts regularly resolve questions of statutory interpretation against defendants without demanding certainty. 258 See, e.g., Shaw v. United States, 137 S. Ct. 462, 469 (2016) (stating that the rule of lenity applies only when there is “a grievous ambiguity or uncertainty in the statute” (internal quotation marks omitted) (quoting Muscarello v. United States, 524 U.S. 125, 138–39 (1998))); Moskal v. United States, 498 U.S. 103, 108 (1990) (explaining that the Court has not “deemed a division of judicial authority automatically sufficient to trigger lenity” (citing United States v. Rodgers, 466 U.S. 475, 484 (1984))).

Finally, requiring only factual certainty—that is, certainty about whether the defendant exclusively admitted (or was convicted of) conduct that constitutes a qualifying offense—would satisfy the practical rationale of Mathis. The Court there reasoned that “[s]tatements of ‘non-elemental fact’ in the records of prior convictions are prone to error precisely because their proof is unnecessary.” 259 Mathis, 136 S. Ct. at 2253 (quoting Descamps v. United States, 133 S. Ct. 2276, 2288–89 (2013)). The Court elaborated, “At trial, and still more at plea hearings, a defendant may have no incentive to contest what does not matter under the law; to the contrary, he ‘may have good reason not to’—or even be precluded from doing so by the court.” 260 Id. Requiring courts to determine with certainty that the defendant exclusively admitted (or that a jury found) qualifying facts ameliorates these concerns about inaccuracy in the record documents. 261 One might respond that Mathis and Descamps established that the record can never establish non-elemental facts with certainty. See United States v. Martinez-Lopez, 864 F.3d 1034, 1045–46 (9th Cir. 2017) (Berzon, J., concurring in part and dissenting in part) (“[O]ver-eager deployment of the modified approach can lead to sentencing enhancements based on information that ‘may be downright wrong,’ and can ‘deprive some defendants of the benefits of their negotiated plea deals.’” (quoting Descamps, 133 S. Ct. at 2289)). Remember, however, that the debate here is over whether certainty is required in determining whether a fact is non-elemental. When a court of appeals finds that the best reading of state law (even if not a certain reading) is that a statute sets out alternative elements, it is far less likely that the record will contain inaccuracies about those facts. Further, the government in Descamps and Mathis had sought inferences based on a defendant’s silence rather than on a specific admission like that at issue in Martinez-Lopez. Compare Mathis, 136 S. Ct. at 2269–70 (Alito, J., dissenting) (describing uncontested allegations in the charging documents), and Descamps, 133 S. Ct. at 2282 (describing the defendant’s failure to object to a prosecutor’s description of the crime), with Martinez-Lopez, 864 F.3d at 1043 (describing the defendant’s admission in a plea colloquy). The Supreme Court, therefore, has not held that legal uncertainty about the means–element question renders it unfair to rely on the defendant’s own admissions.
Allowing a conviction when divisibility is less than certain may seem to raise the same practical concerns as a case in which a statute is plainly indivisible. That is, a defendant might lack incentives to challenge incorrect facts if state law is less than clear about whether they are required for conviction. However, much the same is true of elemental facts that do not affect the maximum sentence. For example, suppose the Ninth Circuit was correct in Martinez-Lopez that sale and offering for sale are distinct elements. See id. Even so, if the difference carries no sentencing consequences, “the defendant may not wish to irk the prosecutor or court by squabbling about superfluous factual allegations.” Descamps, 133 S. Ct. at 2289. Therefore, the additional convictions that this Note would allow to support enhancements are at least no more practically concerning than some convictions that Mathis has already found sufficient. If a future Congress seeks a more straightforward approach, it could key prior-crime enhancements to the length of the prior sentence. See Almanza–Arenas v. Lynch, 815 F.3d 469, 483 (9th Cir. 2016) (Owens, J., concurring) (advocating “a more objective standard, such as the length of the underlying sentence” to determine whether an offence is a crime involving moral turpitude). After all, the length of the sentence is almost guaranteed to be the focus of adversarial testing.
Such certainty could be established by an indictment and plea colloquy (or jury instructions) that “referenc[e] one alternative term to the exclusion of all others.” 262 Mathis, 136 S. Ct. at 2257. The word “exclusion” is critical here. Though this Note concludes that the Martinez-Lopez majority has the right methodological approach, that is not an endorsement of its ultimate result. In fact, as Judge Berzon persuasively notes, the felony complaint against Martinez-Lopez “charged him with ‘the crime of SALE/TRANSPORTATION/ OFFER TO SELL CONTROLLED SUBSTANCE.’” Martinez-Lopez, 864 F.3d at 1055 (Berzon, J., concurring in part and dissenting in part). Therefore, Martinez-Lopez’s admission that he sold cocaine cannot rule out the possibility that his conviction was based only on an offer to sell, which would not be a qualifying predicate. Id. at 1037–38 & n.3 (citing United States v. Rivera-Sanchez, 247 F.3d 905, 909 (9th Cir. 2001) (en banc), superseded on other grounds by U.S. Sentencing Guidelines Manual § 2L1.2 cmt. n.4 (U.S. Sentencing Comm’n 2002)). Given the indictment, if Martinez-Lopez had gone to trial, presumably the government could have had the jury instructed that it need only find that the defendant made an offer to sell. Therefore, this is a case in which the defendant could have been convicted of a qualifying offense, but was not necessarily so convicted. Cf. United States v. Horse Looking, 828 F.3d 744, 749 (8th Cir. 2016) (reaching a similar conclusion). But cf. Mathis, 136 S. Ct. at 2270–71 (Alito, J., dissenting) (imagining a plea colloquy that would exclude conviction based on a non-qualifying means).

C. The Insufficiency of Indictments to Prove What a Defendant Necessarily Admitted

The differing opinions in Gundy highlight that Mathis’s instruction that record documents must “speak plainly” to show that a fact was ele­mental is not self-defining. The panel majority thought the indictments spoke with sufficient clarity that entry of a building was elemental 263 United States v. Gundy, 842 F.3d 1156, 1168–69 (11th Cir. 2016). while Judge Pryor characterized the documents as unclear and questioned whether indictments alone could ever satisfy the demand for certainty. 264 Id. at 1177–78 & n.9 (Pryor, J., dissenting). This section argues that Judge Pryor’s latter suggestion is correct—an indictment, without more, cannot form the basis of applying a prior-conviction enhancement.

In Gundy, the indictments each contained four components that shed some light on his conviction. They (1) alleged “burglary,” (2) cited the Georgia Code provision defining burglary, (3) alleged that Gundy unlawfully entered a business house with intent to commit a theft there­in, and (4) identified the specific location of the alleged theft. 265 Id. at 1168–69 (majority opinion). Gundy then pleaded guilty “[u]pon the foregoing accusation, including each and every charge and count therein contained.” 266 Supplemental Appendix at 8, Gundy, 842 F.3d 1156 (No. 14-12113-CC) (providing Gundy’s original guilty-plea form). The Gundy majority considered the third component of the indictment sufficient indication that the “business house” location was elemental, 267 See Gundy, 842 F.3d at 1170. while Judge Pryor treated it as no different than the fourth element, alleging specific facts not essential to the ultimate conviction. 268 See id. at 1175–78 (Pryor, J., dissenting).

This example highlights the importance of how one treats indict­ments. In the Gundy majority’s view, when a defendant pleads guilty, a sentencing judge can “rely on the indictments, not pattern jury instruc­tions never given.” 269 Id. at 1169 n.10 (majority opinion). Judge Pryor disagreed, saying that indictments do not answer “the determinative question: at trial, what must a Georgia jury find beyond a reasonable doubt to convict the defendant of burglary?” 270 Id. at 1177 (Pryor, J., dissenting) (citing Mathis v. United States, 136 S. Ct. 2243, 2248 (2016)). The resolution of this question is critically important because most prior convictions result not from trials but from guilty pleas. Though Mathis treated the archetypal record documents as an “indictment and correla­tive jury instructions,” 271 Mathis, 136 S. Ct. at 2257. in the vast majority of cases the record will con­tain no such jury instructions. Can the inclusion of facts in an indictment alone “speak plainly” that those facts are elemental?

Under Mathis, the answer to that question should be no. The Court said that the categorical approach focuses solely on elements, and then went on to define elements: “At a trial, they are what the jury must find beyond a reasonable doubt to convict the defendant; and at a plea hear­ing, they are what the defendant necessarily admits when he pleads guilty.” 272 Id. at 2248 (citations omitted). The word “necessarily” is critical. An indictment may include facts that are not necessary to a conviction, facts that would not appear in the jury instructions if the defendant went to trial. Thus, the indictment alone cannot conclusively show that a fact was elemental. Record docu­ments produced from guilty pleas will often show that the defendant actually admitted certain facts but not that the defendant necessarily ad­mitted those facts. To determine which of the admitted facts were ele­mental, the sentencing judge will need to parse state law to predict what facts a jury would have been required to find if the case had gone to trial. This further narrows the availability of the modified categorical ap­proach, but any other reading is hard to square with the text of Mathis and with Apprendi’s bar on judicial fact-finding.


The Supreme Court’s jurisprudence of prior convictions has under­gone a long evolution, in which Mathis is probably just the latest step. Recent cases applying Mathis show that the decision provides sufficient tools to resolve most cases in a nonarbitrary way. In most cases, the fed­eral judiciary can apply its core skills of common-law and statutory inter­pretation to identify whether an alternatively phrased statute lists elements or means. Gray areas remain, however, in courts’ use of the record documents to define their divisibility analysis.

This Note suggests that courts should adopt three interpretive clari­fications of Mathis. First, they should not read its characterization of the means–element distinction as a “threshold inquiry” to require that courts adjudicate the divisibility of a statute when the case could be dis­posed of on the simpler grounds that the record documents do not clearly identify which statutory alternative formed the basis of the de­fendant’s conviction. Second, they should not apply “Taylor’s demand for certainty” to the legal question of divisibility but only to the factual ques­tion of whether the record shows that the defendant necessarily admitted (or was convicted of) a qualifying offense. And third, courts should not treat facts appearing only in the indictments as sufficient proof that a defendant “necessarily admitted” those facts, even if the defendant pleaded guilty to the indictment without qualification.

The Court’s recent decisions have placed increasing pressure on the modified categorical approach. By adopting the constructions suggest­ed here, courts can apply the modified categorical approach in a way that is faithful to Mathis, fair to defendants, and focuses judicial resources into the inquir­ies that most efficiently resolve cases.


Appendix: Circuit Decisions

The following table identifies the cases considered in this Note. The criteria for inclusion were as follows: (1) the opinion cites Mathis, as re­flected in the Westlaw database; (2) the opinion expresses a position on whether a statute is divisible, even if in dicta; (3) the opinion provides some independent reasoning for the divisibility determination that it reaches; and (4) the case was decided prior to October 2017. The third criterion means that the Note does not consider cases in which the court of appeals never conducted a divisibility inquiry because either the government or the defendant had conceded the point or because one party’s position was foreclosed by circuit precedent.

The columns of the table indicate: (1) the name of the case, (2) the court that decided the case, (3) the date of decision, (4) the type of case in which the divisibility issue arose, (5) whether the court found the stat­ute divisible, (6) whether the court imposed the prior-conviction sanc­tion, (7) whether there was a dissent from the decision, (8) whether the court based its divisibility decision in part on state case law, and (9) whether the court based its divisibility decision in part on the record documents from the prior conviction.

Court Date Context Divisible Imposed
Dissent State
Gomez-Perez v. Lynch, 829 F.3d 323 5th Cir. 7/11/16 Immigration No No No Yes No
United States v. Horse Looking, 828 F.3d 744 8th Cir. 7/11/16 18 U.S.C. § 922(g)(9) Yes No No No No
United States v. Lopez–Jacobo, 656 F. App’x 409 10th Cir. 7/22/16 Guidelines Yes Yes No No Yes
United States v. Headbird, 832 F.3d 844 8th Cir. 8/9/16 ACCA No No No Yes No
United States v. Hinkle, 832 F.3d 569 5th Cir. 8/11/16 Guidelines No No No Yes No
United States v. Sheffield, 832 F.3d 296 D.C. Cir. 8/12/16 Guidelines No No No Yes No
Chang-Cruz v. Attorney Gen., 659 F. App’x 114 3d. Cir 8/24/16 Immigration No No No Yes Yes
United States v. Alfaro, 835 F.3d 470 4th Cir. 8/29/16 Guidelines Yes Yes No No Yes
United States v. Lara-Martinez, 836 F.3d 472 5th Cir. 9/6/16 Guidelines Yes Yes No No No
United States v. Edwards, 836 F.3d 831 7th Cir. 9/8/16 Guidelines No No No No Yes
United States v. Fogg, 836 F.3d 951 8th Cir. 9/8/16 ACCA Yes Yes Yes Yes No
Spaho v. Attorney Gen., 837 F.3d 1172 11th Cir. 9/19/16 Immigration Yes Yes Yes Yes No
United States v. Howell, 838 F.3d 489 5th Cir. 9/22/16 Guidelines No Yes No Yes No
United States v. Uribe, 838 F.3d 667 5th Cir. 10/3/16 Guidelines Yes Yes No Yes No
United States v. Bryant, 669 F. App’x 238 5th Cir. 10/4/16 Guidelines Yes Yes No No Yes
Singh v. Attorney Gen., 839 F.3d 273 3d Cir. 10/6/16 Immigration Yes No No Yes Yes
Unites States v. De La O-Gallegos, 663 F. App’x 827 11th Cir. 10/7/16 Guidelines Yes Yes No No No
United States v. Maldonado-Palma, 839 F.3d 1244 10th Cir. 10/25/16 Guidelines Yes Yes No No No
United States v. Ritchey, 840 F.3d 310 6th Cir. 10/26/16 ACCA No No No Yes No
United States v. Haney, 840 F.3d 472 7th Cir. 10/27/16 ACCA No No No No No
In re McComb, 691 F. App’x 819 6th Cir. 11/3/16 Guidelines Yes Yes No No No
United States v. Redrick, 841 F.3d 478 D.C. Cir. 11/8/16 ACCA Yes Yes No Yes No
United States v. Henderson, 841 F.3d 623 3d Cir. 11/8/16 ACCA Yes Yes No Yes Yes
United States v. McFee, 842 F.3d 572 8th Cir. 11/17/16 ACCA No No No Yes Yes
Garcia v. Lynch, 670 F. App’x 647 9th Cir. 11/18/16 Immigration Yes Yes No No No
United States v. Cardena, 842 F.3d 959 7th Cir. 11/18/16 18 U.S.C. § 924(c) Yes Yes No Yes Yes
United States v. Esprit, 841 F.3d 1235 11th Cir. 11/21/16 ACCA No No No Yes No
United States v. Gundy, 842 F.3d 1156 11th Cir. 11/23/16 ACCA Yes Yes Yes Yes Yes
United States v. Tavares, 843 F.3d 1 1st Cir. 12/1/16 Guidelines Yes Remand No Yes No
United States v. Taylor, 672 F. App’x 860 10th Cir. 12/6/16 ACCA No Yes No No No
United States v. Rocha-Alvarado, 843 F.3d 802 9th Cir. 12/12/16 Guidelines Yes Yes No No No
United States v. Taylor, 843 F.3d 1215 10th Cir. 12/12/16 Guidelines Yes Yes No No No
United States v. Bernel-Aveja, 844 F.3d 206 5th Cir. 12/13/16 Immigration No No No No No
United States v. Sykes, 844 F.3d 712 8th Cir. 12/21/16 ACCA Yes Yes No No Yes
United States v. Starks, 674 F. App’x 580 8th Cir. 12/28/16 Guidelines Yes Yes No No No
United States v. Parrow, 844 F.3d 801 8th Cir. 12/30/16 Guidelines No Yes No Yes No
United States v. Harris, 844 F.3d 1260 10th Cir. 1/4/17 ACCA Yes Yes Yes No No
Flores-Larrazola v. Lynch, 854 F.3d 732 5th Cir. 1/6/17 Immigration Yes Yes No Yes No
United States v. Winston, 845 F.3d 876 8th Cir. 1/10/17 ACCA Yes Yes No No No
Ibanez-Beltran v. Lynch, 858 F.3d 294 5th Cir. 1/11/17 Immigration Yes Yes No Yes Yes
United States v. Garcia-Martinez, 845 F.3d 1126 11th Cir. 1/11/17 Guidelines No No No Yes No
United States v. Tanksley, 848 F.3d 347 5th Cir. 1/18/17 Guidelines No No No Yes No
United States v. Hertz, 673 F. App’x 606 8th Cir. 1/25/17 ACCA No No No No No
United States v. Solano-Hernandez, 847 F.3d 170 5th Cir. 1/26/17 Guidelines Yes Yes No Yes No
Sandoval v. Sessions, 866 F.3d 986 9th Cir. 1/27/17 Immigration No No No No No
Swaby v. Yates, 847 F.3d 62 1st Cir. 1/30/17 Immigration Yes Yes No Yes Yes
United States v. Mendez-Henriquez, 847 F.3d 214 5th Cir. 1/30/17 Guidelines Yes Yes Yes Yes Yes
United States v. Dozier, 848 F.3d 180 4th Cir. 1/30/17 Guidelines Yes Yes No No Yes
United States v. Steiner, 847 F.3d 103 3d Cir. 2/1/17 Guidelines No No No Yes Yes
Gatson v. United States, 2017 WL 3224851 11th Cir. 2/1/17 ACCA Yes Yes No Yes Yes
United States v. Taylor, 848 F.3d 476 1st Cir. 2/8/17 ACCA Yes Yes No No No
United States v. Lobaton-Andrade, 861 F.3d 538 5th Cir. 2/9/17 Guidelines No No No Yes No
United States v. Delgado-Sánchez, 849 F.3d 1 1st Cir. 2/17/17 Guidelines Yes Yes No No No
United States v. Alexander, 680 F. App’x 388 6th Cir. 2/22/17 Guidelines No No No Yes No
United States v. McArthur, 850 F.3d 925 8th Cir. 2/23/17 ACCA No No No Yes No
United States v. Irons, 849 F.3d 743 8th Cir. 2/27/17 ACCA Yes Yes No No No
United States v. Gooch, 850 F.3d 285 6th Cir. 3/2/17 18 U.S.C. § 924(c) Yes Yes No Yes Yes
United States v. Gonzalez-Lince, 678 F. App’x 270 5th Cir. 3/8/17 Guidelines Yes Yes Yes No No
Chavez-Alvarez v. Attorney Gen., 850 F.3d 583 3d Cir. 3/9/17 Immigration No No No Yes No
United States v. Hudson, 851 F.3d 807 8th Cir. 3/21/17 Guidelines Yes Yes No Yes No
United States v. Lynn, 851 F.3d 786 7th Cir. 3/24/17 Guidelines Yes Yes No Yes No
United States v. Titties, 852 F.3d 1257 10th Cir. 3/24/17 ACCA No No Yes Yes Yes
United States v. Faust (Mass. Resisting Arrest), 853 F.3d 39 1st Cir. 4/5/17 ACCA No No No No Yes
United States v. Faust (Mass. Assault and Battery on a Police Officer), 853 F.3d 39 1st Cir. 4/5/17 ACCA Yes Remand No Yes Yes
United States v. Hall, 684 F. App’x 333 4th Cir. 4/7/17 ACCA No No No No No
United States v. Tibbs, 685 F. App’x. 456 6th Cir. 4/10/17 Guidelines Yes Yes No Yes No
United States v. Ama, 684 F. App’x. 736 10th Cir. 4/11/17 ACCA No No No No No
Castendet-Lewis v. Sessions, 855 F.3d 253 4th Cir. 4/25/17 Immigration No No No Yes Yes
United States v. Martinez-Rodriguez, 857 F.3d 282 5th Cir. 5/12/17 Guidelines No No No Yes No
United States v. Harrison, 691 F. App’x. 440 9th Cir. 5/24/17 ACCA No No No Yes No
Diego v. Sessions, 857 F.3d 1005 9th Cir. 5/26/17 Asylum Yes Yes No Yes Yes
Lofties v. United States, 694 F. App’x. 996 6th Cir. 6/1/17 ACCA Yes Yes No No No
Moring v. United States, 2017 WL 4574491 6th Cir. 6/8/17 Guidelines Yes Yes No No No
Ginter v. United States, 2017 WL 4570519 6th Cir. 6/12/17 ACCA No No No No No
United States v. Rogers, 696 F. App’x 878 10th Cir. 6/13/17 Guidelines Yes Yes No Yes No
Harbin v. Sessions, 860 F.3d 58 2d Cir. 6/21/17 Immigration No No No Yes No
United States v. Goodson, 700 F. App’x 417 6th Cir. 6/26/17 Guidelines Yes Yes No Yes No
United States v. Stitt, 860 F.3d 854 6th Cir. 6/27/17 ACCA No No Yes No No
United States v. Perez-Silvan, 861 F.3d 935 9th Cir. 6/28/17 Immigration Yes Yes No Yes No
United States v. Calvillo-Palacios (Texas Simple Assault), 860 F.3d 1285 9th Cir. 6/28/17 Guidelines Yes Yes No Yes No
United States v. Calvillo-Palacios (Texas Aggravated Assault), 860 F.3d 1285 9th Cir. 6/28/17 Guidelines No Yes No Yes No
United States v. Reyes-Ochoa, 861 F.3d 582 5th Cir. 6/30/17 Guidelines No No No Yes No
United States v. Ochoa, 861 F.3d 1010 9th Cir. 7/3/17 Immigration No No No Yes Yes
Gordon v. Attorney Gen., 861 F.3d 1314 11th Cir. 7/10/17 Immigration Yes No No Yes No
United States v. McMillan, 863 F.3d 1053 8th Cir. 7/24/17 Guidelines No No No Yes No
United States v. Maldonado, 864 F.3d 893 8th Cir. 7/25/17 Guidelines Yes Yes No No No
United States v. Diaz, 865 F.3d 168 4th Cir. 7/26/17 Mandatory Victims Restitution Act No No No Yes No
United States v. Martinez-Lopez (Controlled-Substance Requirement), 864 F.3d 1034 9th Cir. 7/28/17 Guidelines Yes Yes No Yes No
United States v. Martinez-Lopez (Actus Reus Requirement), 864 F.3d 1034 9th Cir. 7/28/17 Guidelines Yes Yes Yes Yes No
United States v. Enoch, 865 F.3d 575 7th Cir. 7/28/17 18 U.S.C. § 924(c) Yes Yes No No No
United States v. Reyes, 866 F.3d 316 5th Cir. 8/1/17 Guidelines Yes Yes Yes Yes Yes
United States v. Madkins, 866 F.3d 1136 10th Cir. 8/8/17 Guidelines Yes No No No No
United States v. Pam, 867 F.3d 1191 10th Cir. 8/15/17 ACCA Yes Yes No No Yes
United States v. Burtons, 696 F. App’x 372 10th Cir. 8/17/17 ACCA Yes Yes No Yes No
United States v. Herrera-Serrano, 703 F. App’x 342 5th Cir. 8/21/17 Guidelines Yes Yes No Yes No
United States v. Perlaza-Ortiz, 869 F.3d 375 5th Cir. 8/23/17 Guidelines No No No Yes Yes
Marinelarena v. Sessions, 869 F.3d 780 9th Cir. 8/23/17 Immigration Yes Yes Yes Yes No
United States v. Robinson, 869 F.3d 933 9th Cir. 8/25/17 Guidelines No No No Yes No
United States v. Mata, 869 F.3d 640 8th Cir. 8/25/17 ACCA Yes Yes No Yes No
United States v. Ocampo-Estrada, 873 F.3d 661 9th Cir. 8/29/17 21 U.S.C. § 841(b)(1)(A) Yes No No Yes No
United States v. Reyes, 697 F. App’x 519 9th Cir. 9/8/17 Guidelines Yes Yes No No No
Laryea v. Sessions, 871 F.3d 337 5th Cir. 9/12/17 Immigration Yes No No No No
Totals 60% 56% 12% 62% 26%